# tinyml

TinyML is a multidisciplinary field enabling machine-learning algorithms on extremely low-power embedded devices, including the hardware and software needed for deployment.

This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles.

## Hypnos: A Dual-Phase End-to-End Platform for Scalable Continual Learning at the Edge

DevFeed: [Hypnos: A Dual-Phase End-to-End Platform for Scalable Continual Learning at the Edge](<https://devfeed.tech/articles/hypnos-a-dual-phase-end-to-end-platform-for-scalable-continual-learning-at-the-edge-142817.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.24637>)

Author: Guilherme Guedes, Marco Apolinario Lainez, Charlotte Frenkel

Published: 2026-11-19T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [tinyml](<https://devfeed.tech/topics/tinyml.md>), [On-device AI](<https://devfeed.tech/topics/on-device-ai.md>), [fpga](<https://devfeed.tech/topics/fpga.md>)

Tags: [backpropagation](<https://devfeed.tech/tags/backpropagation.md>), [charger](<https://devfeed.tech/tags/charger.md>), [charging](<https://devfeed.tech/tags/charging.md>), [continual-learning](<https://devfeed.tech/tags/continual-learning.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-ar](<https://devfeed.tech/tags/cs-ar.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [edge](<https://devfeed.tech/tags/edge.md>), [fpga](<https://devfeed.tech/tags/fpga.md>), [gradient](<https://devfeed.tech/tags/gradient.md>), [inference](<https://devfeed.tech/tags/inference.md>), [latency](<https://devfeed.tech/tags/latency.md>), [mcu](<https://devfeed.tech/tags/mcu.md>), [memory](<https://devfeed.tech/tags/memory.md>), [online](<https://devfeed.tech/tags/online.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [platform](<https://devfeed.tech/tags/platform.md>), [real-world](<https://devfeed.tech/tags/real-world.md>), [risc-v](<https://devfeed.tech/tags/risc-v.md>), [sample](<https://devfeed.tech/tags/sample.md>), [sensor](<https://devfeed.tech/tags/sensor.md>), [sleep](<https://devfeed.tech/tags/sleep.md>), [system-on-chip](<https://devfeed.tech/tags/system-on-chip.md>)

### AI overview

Hypnos splits edge continual learning between a low-resource "wake" phase and a better-resourced "sleep" phase. A RISC-V microcontroller handles inference and lightweight, gradient-free prototype updates as sensor samples arrive; a smart charger uses rehearsal-based gradient updates for consolidation when more energy and memory are available. On CORe50 NICv2-391, deployed on a ZCU104 FPGA and synthesized in 28-nm technology, it reports 78.6%-81.6% accuracy and 7.79-10.54 seconds per sample, depending on activation precision. Compared with quantized latent replay, it reports 3.8x fewer learning steps and 22x less learning compute, with higher accuracy and comparable latency.

### Source excerpt

arXiv:2611.24637v1 Announce Type: cross Abstract: Continual learning (CL) enables neural networks to adapt to non-stationary data streams by mitigating catastrophic forgetting, offering a path to robust, private, and adaptable edge devices. Yet, porting CL to real-world edge devices remains challenging for two reasons: (i) online operation, where sensor data arrives one sample at a time and the device must adapt on-the-fly, and (ii) the low footprint of edge devices, which makes traditional error backpropagation training prohibitive. Inspired by the brain's wake and sleep cycles, the former for rapid adaptation with minimal resources and the latter for long-term consolidation, we introduce Hypnos, an open-source dual-phase platform for efficient CL at the edge. In the wake phase (e.g., daily operation), Hypnos relies on a low-footprint system-on-chip with a RISC-V microcontroller unit (MCU) targeting efficient inference and supporting lightweight gradient-free prototype updates. In the sleep phase (e.g., night-time charging), where wall-plug energy and larger memory budgets are available, a smart charger performs rehearsal-based gradient updates for consolidation. Deployed on a ZCU104 field-programmable gate array (FPGA) and synthesized in a 28-nm process, Hypnos is benchmarked on the CORe50 NICv2-391 protocol. It sets a new state-of-the-art accuracy of 78.6%-81.6% for edge CL systems, with an on-device latency of 7.79s-10.54s per sample depending on activation precision, and requires $3.8\times$ fewer learning steps and $22\times$ less learning compute than quantized latent replay at higher accuracy and comparable latency.

## TinyMyo: a Tiny Foundation Model for Flexible EMG Signal Processing at the Edge

DevFeed: [TinyMyo: a Tiny Foundation Model for Flexible EMG Signal Processing at the Edge](<https://devfeed.tech/articles/tinymyo-a-tiny-foundation-model-for-flexible-emg-signal-processing-at-the-edge-142500.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2512.15729>)

Author: Matteo Fasulo, Giusy Spacone, Thorir Mar Ingolfsson, Yawei Li, Luca Benini, Andrea Cossettini

Published: 2026-11-19T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>), [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [AI Foundation Models](<https://devfeed.tech/topics/ai-foundation-models.md>), [tinyml](<https://devfeed.tech/topics/tinyml.md>), [Image processing](<https://devfeed.tech/topics/image-processing.md>), [Model Development](<https://devfeed.tech/topics/model-development.md>), [model-deployment](<https://devfeed.tech/topics/model-deployment.md>), [Representation learning](<https://devfeed.tech/topics/representation-learning.md>)

Tags: [classification](<https://devfeed.tech/tags/classification.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-hc](<https://devfeed.tech/tags/cs-hc.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [datasets](<https://devfeed.tech/tags/datasets.md>), [decoding](<https://devfeed.tech/tags/decoding.md>), [deployment](<https://devfeed.tech/tags/deployment.md>), [edge](<https://devfeed.tech/tags/edge.md>), [eess-sp](<https://devfeed.tech/tags/eess-sp.md>), [flexible](<https://devfeed.tech/tags/flexible.md>), [foundation](<https://devfeed.tech/tags/foundation.md>), [foundation-model](<https://devfeed.tech/tags/foundation-model.md>), [inference](<https://devfeed.tech/tags/inference.md>), [interaction](<https://devfeed.tech/tags/interaction.md>), [low-power](<https://devfeed.tech/tags/low-power.md>), [microcontroller](<https://devfeed.tech/tags/microcontroller.md>), [motor](<https://devfeed.tech/tags/motor.md>), [on-device](<https://devfeed.tech/tags/on-device.md>), [signal-processing](<https://devfeed.tech/tags/signal-processing.md>), [speech](<https://devfeed.tech/tags/speech.md>), [speech-recognition](<https://devfeed.tech/tags/speech-recognition.md>), [transformer](<https://devfeed.tech/tags/transformer.md>)

### AI overview

TinyMyo is a 3.6-million-parameter Transformer encoder trained on public EMG datasets with self-supervised masked reconstruction. The authors evaluated it on hand gesture classification, hand kinematic regression, speech production, and speech recognition, adapting it with task-specific prediction heads. For gesture classification, they report 87.98% accuracy on NinaPro DB5, 97.1% on UCI-EMG, and 96.6% on EPN-612, outperforming previously reported EMG foundation models. On a GAP9 ultra-low-power microcontroller, inference took 0.785 seconds and used 44.91 mJ per inference at 57.18 mW average power. A smaller 1.9-million-parameter variant processed 5-second and 1-second windows in 0.496 and 0.089 seconds, respectively. These abstract results indicate support for multiple tasks and edge deployment; the excerpt does not provide further methodological detail.

### Source excerpt

arXiv:2512.15729v3 Announce Type: replace-cross Abstract: Objective: Surface electromyography (EMG) is widely used in biomechanics, rehabilitation, prosthetic control, and human-machine interfaces. Generalization across subjects, recording systems, and acquisition protocols remains challenging. Existing EMG foundation models (FMs) have limited support for multiple downstream tasks and deployment on low-power devices. Methods: We introduce TinyMyo, a 3.6-million-parameter Transformer encoder pre-trained on public EMG datasets using self-supervised masked reconstruction. TinyMyo supports multiple tasks with minimal changes to the task-specific prediction head. Results: We evaluated TinyMyo on hand gesture classification, hand kinematic regression, speech production, and speech recognition. For gesture classification, TinyMyo outperformed previously reported EMG FMs, achieving accuracies of 87.98% on NinaPro DB5, 97.1% on UCI-EMG, and 96.6% on EPN-612. On the GAP9 ultra-low-power microcontroller, TinyMyo achieved an inference time of 0.785 s and consumed 44.91 mJ per inference at an average power of 57.18 mW. A 1.9-million-parameter variant processed 5 s and 1 s input windows in 0.496 s and 0.089 s, respectively. Conclusion: TinyMyo demonstrates that a compact, self-supervised EMG encoder can support multiple downstream tasks and run on a low-power microcontroller. Significance: To our knowledge, this is the first deployment of an EMG FM on an ultra-low-power microcontroller, opening a path to on-device motor intent decoding, neuromuscular assessment, and human-machine interaction.

## U-blox brings WiFi 6E and edge AI to low-power IoT modules

DevFeed: [U-blox brings WiFi 6E and edge AI to low-power IoT modules](<https://devfeed.tech/articles/u-blox-brings-wifi-6e-and-edge-ai-to-low-power-iot-modules-141271.md>)

Original publisher: [Read original article](<https://blog.adafruit.com/2026/11/18/u-blox-brings-wifi-6e-and-edge-ai-to-low-power-iot-modules/>)

Author: Anne Barela

Published: 2026-11-18T22:22:45Z

Content type: news

Language: en

Sources: [Adafruit Industries](<https://devfeed.tech/sources/adafruit-industries-makers-hackers-artists-designers-and-engineers.md>)

Topics: [Internet of things](<https://devfeed.tech/topics/iot.md>), [Bluetooth LE](<https://devfeed.tech/topics/bluetooth-le.md>), [zigbee](<https://devfeed.tech/topics/zigbee.md>), [tinyml](<https://devfeed.tech/topics/tinyml.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [bluetooth](<https://devfeed.tech/tags/bluetooth.md>), [components-parts](<https://devfeed.tech/tags/components-parts.md>), [edge](<https://devfeed.tech/tags/edge.md>), [edge-ai](<https://devfeed.tech/tags/edge-ai.md>), [electronics](<https://devfeed.tech/tags/electronics.md>), [internet-of-things](<https://devfeed.tech/tags/internet-of-things.md>), [iot](<https://devfeed.tech/tags/iot.md>), [low-power](<https://devfeed.tech/tags/low-power.md>), [microcontrollers](<https://devfeed.tech/tags/microcontrollers.md>), [modules](<https://devfeed.tech/tags/modules.md>), [nrf7120](<https://devfeed.tech/tags/nrf7120.md>), [u-blox](<https://devfeed.tech/tags/u-blox.md>), [wifi](<https://devfeed.tech/tags/wifi.md>), [wifi-6e](<https://devfeed.tech/tags/wifi-6e.md>)

### AI overview

U-blox's ALMA-W1 modules combine WiFi 6, Bluetooth LE and, in wireless MCU versions, Thread and Zigbee for low-power IoT devices. WiFi 6E models add 6 GHz; u-blox reports 8 µA connected standby at 5 GHz using DTIM10, while receive current is 10.3-11.7 mA and transmit current reaches 241 mA at full output. Wireless MCU models include a 256 MHz Arm Cortex-M33, 256 MHz RISC-V coprocessor, 1 MB RAM and 63 GPIOs. Tri-band ALMA-W111 and ALMA-W116 modules add Nordic's Axon NPU for supported on-module AI workloads; u-blox claims up to 15x the speed of the main MCU. The smallest package measures 10.4 x 11.2 x 1.9 mm. Samples are going to select customers, and certifications are pending.

### Source excerpt

All About Circuits covered u-blox's new ALMA-W1 series, a family of small wireless modules built around Nordic Semiconductor's nRF7120 and nRF7120E chips. They support WiFi 6 on 2.4 GHz and 5 GHz, with WiFi 6E models adding the 6 GHz band, plus Bluetooth LE. The wireless MCU versions also handle Thread and Zigbee. They're aimed [...]

## Real-time fall detection based on vision for low-power edge platforms

DevFeed: [Real-time fall detection based on vision for low-power edge platforms](<https://devfeed.tech/articles/real-time-fall-detection-based-on-vision-for-low-power-edge-platforms-139790.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2607.12909>)

Author: Wenjun Xia, Zhicheng Peng, Haopeng Li, Zhengdi Zhang

Published: 2026-11-18T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [tinyml](<https://devfeed.tech/topics/tinyml.md>), [real-time](<https://devfeed.tech/topics/real-time.md>)

Tags: [continuous](<https://devfeed.tech/tags/continuous.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [detection](<https://devfeed.tech/tags/detection.md>), [devices](<https://devfeed.tech/tags/devices.md>), [early-warning](<https://devfeed.tech/tags/early-warning.md>), [edge](<https://devfeed.tech/tags/edge.md>), [edge-devices](<https://devfeed.tech/tags/edge-devices.md>), [elderly](<https://devfeed.tech/tags/elderly.md>), [elderly-care](<https://devfeed.tech/tags/elderly-care.md>), [fall](<https://devfeed.tech/tags/fall.md>), [fall-detection](<https://devfeed.tech/tags/fall-detection.md>), [framework](<https://devfeed.tech/tags/framework.md>), [inference](<https://devfeed.tech/tags/inference.md>), [low-power](<https://devfeed.tech/tags/low-power.md>), [physics](<https://devfeed.tech/tags/physics.md>), [physics-informed](<https://devfeed.tech/tags/physics-informed.md>), [prediction](<https://devfeed.tech/tags/prediction.md>), [q-bio-nc](<https://devfeed.tech/tags/q-bio-nc.md>), [real-time](<https://devfeed.tech/tags/real-time.md>), [vision](<https://devfeed.tech/tags/vision.md>)

### AI overview

The paper proposes vision-based fall detection as a loss of stability in a coupled human support system. Its dual-LTC architecture models Center-of-Mass motion and Base-of-Support adjustment, connects those subsystems with a learnable coupling module, and uses a Stability Manifold classifier with Lyapunov-inspired metrics; counterfactual trajectories and Time-to-Collision estimates are intended to assess irreversibility and provide early warning. The authors report a network under 50K parameters for real-time inference on low-power edge devices and competitive accuracy with greater physical interpretability than conventional CNN-RNN pipelines. Although designed for Normal, Falling and Fallen states, this preliminary study validates only Normal versus Falling; full three-state temporal transitions remain future work.

### Source excerpt

arXiv:2607.12909v2 Announce Type: replace-cross Abstract: Falling detection is vital for elderly care and intelligent surveillance; however, prevailing vision-based approaches predominantly frame it as static pose classification or discrete temporal pattern matching, fundamentally overlooking the instability dynamics of the human support system. This paper proposes a physics-informed falling detection framework that recasts falling as a stability-loss event in a coupled dynamical system. We introduce a novel dual-LTC architecture comprising a Center-of-Mass (CoM) subsystem and a Base-of-Support (BoS) subsystem, both instantiated as Liquid Time-Constant (LTC) neural networks to continuously model inertial trajectory evolution and ground-contact adjustment through adaptive time constants, Physical interpretability of falling motion. A learnable coupling module emulates physical interaction between the two subsystems, while a Stability Manifold classifier operates in the joint latent space to detect boundary crossing via Lyapunov-inspired stability metrics. Complementary counterfactual trajectory projection and Time-to-Collision (TTC) estimation further enable irreversibility assessment and early warning. The architecture is designed to support a three-state prediction paradigm (Normal, Falling, Fallen); in this preliminary study, we validate the core stability discrimination capability on a two-class dataset (Normal vs. Falling), leaving the full three-state temporal transition to future work. Unlike conventional CNN--RNN pipelines, the proposed formulation encodes continuous-time mechanical inertia, yielding a sub-50K-parameter network capable of real-time inference on resource-constrained edge devices. Extensive experiments demonstrate competitive accuracy with superior physical interpretability, validating its efficacy for low-compute visual fall detection.

## WiFi, Bluetooth LE, Thread, and on-device AI on a single chip, with existing software and tools, built for billions of intelligent devices that run for years on a battery

DevFeed: [WiFi, Bluetooth LE, Thread, and on-device AI on a single chip, with existing software and tools, built for billions of intelligent devices that run for years on a battery](<https://devfeed.tech/articles/wifi-bluetooth-le-thread-and-on-device-ai-on-a-single-chip-with-existing-software-and-tools-built-for-billions-of-intelligent-devices-that-run-for-years-on-a-battery-137832.md>)

Original publisher: [Read original article](<https://blog.adafruit.com/2026/11/17/wifi-bluetooth-le-thread-and-on-device-ai-on-a-single-chip-with-existing-software-and-tools-built-for-billions-of-intelligent-devices-that-run-for-years-on-a-battery/>)

Author: Anne Barela

Published: 2026-11-17T15:29:42Z

Content type: release

Language: en

Sources: [Adafruit Industries](<https://devfeed.tech/sources/adafruit-industries-makers-hackers-artists-designers-and-engineers.md>)

Topics: [tinyml](<https://devfeed.tech/topics/tinyml.md>), [Bluetooth LE](<https://devfeed.tech/topics/bluetooth-le.md>), [Local AI](<https://devfeed.tech/topics/local-ai.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [battery](<https://devfeed.tech/tags/battery.md>), [bluetooth](<https://devfeed.tech/tags/bluetooth.md>), [bluetooth-le](<https://devfeed.tech/tags/bluetooth-le.md>), [chip](<https://devfeed.tech/tags/chip.md>), [components-parts](<https://devfeed.tech/tags/components-parts.md>), [edge](<https://devfeed.tech/tags/edge.md>), [edge-ai](<https://devfeed.tech/tags/edge-ai.md>), [ee](<https://devfeed.tech/tags/ee.md>), [electronics](<https://devfeed.tech/tags/electronics.md>), [energy](<https://devfeed.tech/tags/energy.md>), [internet-of-things](<https://devfeed.tech/tags/internet-of-things.md>), [microcontrollers](<https://devfeed.tech/tags/microcontrollers.md>), [nrf71](<https://devfeed.tech/tags/nrf71.md>), [on-device](<https://devfeed.tech/tags/on-device.md>), [on-device-ai](<https://devfeed.tech/tags/on-device-ai.md>), [power](<https://devfeed.tech/tags/power.md>), [semiconductor](<https://devfeed.tech/tags/semiconductor.md>), [silicon](<https://devfeed.tech/tags/silicon.md>), [standby](<https://devfeed.tech/tags/standby.md>), [system-on-chip](<https://devfeed.tech/tags/system-on-chip.md>), [thread](<https://devfeed.tech/tags/thread.md>), [wifi](<https://devfeed.tech/tags/wifi.md>), [wifi-6](<https://devfeed.tech/tags/wifi-6.md>)

### AI overview

Nordic Semiconductor's nRF71 Series combines WiFi, Bluetooth LE, Thread, and on-device AI on one chip for battery-powered devices. It has separate WiFi and 2.4 GHz multiprotocol radios that can receive simultaneously, allowing a device to maintain a WiFi cloud link while listening for Bluetooth LE or Thread traffic. Nordic claims standby power up to 20 times lower than its nRF70 Series WiFi parts. The Axon NPU runs models 10-17 times faster than a 256 MHz Arm Cortex-M33 alone; cited uses include keyword spotting, anomaly detection, and image classification. The nRF7120 offers dual-band WiFi 6, while the nRF7120E adds tri-band WiFi 6E and the Axon NPU. Both include 4 MB non-volatile memory and 1 MB user RAM, and are sampling with selected customers.

### Source excerpt

Nordic Semiconductor has announced the nRF71 Series, a line of system-on-chip parts that put WiFi, Bluetooth Low Energy, Thread, and edge AI on one piece of silicon. Nordic says standby power is up to 20 times lower than its current nRF70 Series WiFi parts, which is the main barrier for battery-powered WiFi designs like smart [...]

## Cyber-Secure Wearable, Ultra-Low-Power Networked Sensors for Soldier Health Monitoring: Insights from the CAPSARII project

DevFeed: [Cyber-Secure Wearable, Ultra-Low-Power Networked Sensors for Soldier Health Monitoring: Insights from the CAPSARII project](<https://devfeed.tech/articles/cyber-secure-wearable-ultra-low-power-networked-sensors-for-soldier-health-monitoring-insights-from-the-capsarii-project-137545.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.20187>)

Author: Luciano Bozzi, Christian Celidonio, Umberto Nuzzi, Massimo Biagini, Stefano Cherubin, Asbj{\\o}rn Djupdal, Tor Andre Haugdahl, Andrea Aliverti, Alessandra Angelucci, Giovanni Agosta, Gerardo Pelosi, Pa

Published: 2026-11-17T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>), [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>), [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [tinyml](<https://devfeed.tech/topics/tinyml.md>), [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [Post-training optimization](<https://devfeed.tech/topics/post-training-optimization.md>), [Cryptography](<https://devfeed.tech/topics/cryptography.md>), [quantization](<https://devfeed.tech/topics/quantization.md>), [Feature Engineering](<https://devfeed.tech/topics/feature-engineering.md>)

Tags: [abstract](<https://devfeed.tech/tags/abstract.md>), [accelerators](<https://devfeed.tech/tags/accelerators.md>), [ai](<https://devfeed.tech/tags/ai.md>), [analysis](<https://devfeed.tech/tags/analysis.md>), [analytics](<https://devfeed.tech/tags/analytics.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [area](<https://devfeed.tech/tags/area.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [ban](<https://devfeed.tech/tags/ban.md>), [based](<https://devfeed.tech/tags/based.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-ar](<https://devfeed.tech/tags/cs-ar.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [cs-et](<https://devfeed.tech/tags/cs-et.md>), [cs-hc](<https://devfeed.tech/tags/cs-hc.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cyber](<https://devfeed.tech/tags/cyber.md>), [health](<https://devfeed.tech/tags/health.md>), [health-monitoring](<https://devfeed.tech/tags/health-monitoring.md>), [insights](<https://devfeed.tech/tags/insights.md>), [low-power](<https://devfeed.tech/tags/low-power.md>), [monitoring](<https://devfeed.tech/tags/monitoring.md>), [power](<https://devfeed.tech/tags/power.md>), [project](<https://devfeed.tech/tags/project.md>), [sensors](<https://devfeed.tech/tags/sensors.md>), [wearable](<https://devfeed.tech/tags/wearable.md>)

### AI overview

CAPSARII outlines a secure, low-power wearable and Internet of Battlefield Things framework for monitoring military personnel. It combines smart-textile sensors, body and squad mesh networking, AI on secure RISC-V microcontrollers, and cloud or federated analytics; pilot fatigue and harsh-environment operations guide its use cases and interfaces. On a public wearable dataset, XGBoost reaches accuracy 0.901±0.023 and F1 0.920±0.019, while a feature-based Transformer records the highest AUC, 0.963±0.025. The project uses hardware/software co-design, pruning, quantization, distillation, and precision tuning for constrained edge devices, and investigates Processing-in-Memory. Remaining challenges include validating signal quality and textile durability, maintaining low-latency BAN/SAN communications amid mobility and interference, and securing constrained devices with robust identity and side-channel resilience.

### Source excerpt

arXiv:2611.20187v1 Announce Type: cross Abstract: The CAPSARII project advances a cyber-secure, ultra-low-power wearable platform and Internet of Battlefield Things (IoBT) framework to support real-time monitoring of military personnel and data-driven decision support. The architecture integrates smart-textile sensing, a body/squad area mesh, edge-deployed AI on secure RISC-V microcontrollers, and cloud/federated analytics for cross-mission learning and comparative studies. Two representative use cases--pilot fatigue and warfighters operating in harsh environments--guide requirements, validation, and interface design (including HUD and haptic feedback). As an initial step toward embedded models, we benchmark stress classification on a public wearable dataset: tree-ensemble baselines achieve state-of-the-art performance for accuracy and F1 (e.g., XGBoost: $\mathrm{Accuracy}=0.901\!\pm\!0.023$, $\mathrm{F1}=0.920\!\pm\!0.019$), while a feature-based Transformer attains the highest AUC ($0.963\!\pm\!0.025$), informing feature-engineering and compression choices for edge deployment. Building on these results, CAPSARII employs HW/SW co-design, model compression (pruning, quantization, distillation), and precision tuning to meet stringent energy and memory budgets; it further investigates Processing-in-Memory accelerators for future efficiency gains. A structured gap analysis highlights the remaining challenges: establishing validation standards for physiological signal quality and textile durability; achieving low-latency, scalable BAN/SAN communications under mobility and interference; and deploying hybrid lightweight/post-quantum cryptography with robust device identity and side-channel resilience on constrained endpoints. By consolidating these elements into a coherent, secure IoBT stack, CAPSARII delineates a practical roadmap from laboratory baselines to mission-ready soldier monitoring in both training and operational scenarios.

## What to Expect at the Zephyr Project Meetup (November 18, 2026) - Hagenberg, Austria

DevFeed: [What to Expect at the Zephyr Project Meetup (November 18, 2026) - Hagenberg, Austria](<https://devfeed.tech/articles/what-to-expect-at-the-zephyr-project-meetup-november-18-2026-hagenberg-austria-137116.md>)

Original publisher: [Read original article](<https://zephyrproject.org/what-to-expect-at-the-zephyr-project-meetup-november-18-2026-hagenberg-austria/>)

Author: Susan Remmert

Published: 2026-11-16T21:25:01Z

Content type: news

Language: en

Sources: [Zephyr Project](<https://devfeed.tech/sources/zephyr-project-2.md>)

Topics: [Zephyr RTOS](<https://devfeed.tech/topics/zephyr-rtos.md>), [tinyml](<https://devfeed.tech/topics/tinyml.md>), [Learning](<https://devfeed.tech/topics/learning.md>), [Open Source](<https://devfeed.tech/topics/open-source.md>)

Tags: [2026](<https://devfeed.tech/tags/2026.md>), [austria](<https://devfeed.tech/tags/austria.md>), [beginner](<https://devfeed.tech/tags/beginner.md>), [blog](<https://devfeed.tech/tags/blog.md>), [bluetooth](<https://devfeed.tech/tags/bluetooth.md>), [bluetooth-le](<https://devfeed.tech/tags/bluetooth-le.md>), [board](<https://devfeed.tech/tags/board.md>), [boot](<https://devfeed.tech/tags/boot.md>), [capabilities](<https://devfeed.tech/tags/capabilities.md>), [community](<https://devfeed.tech/tags/community.md>), [embedded](<https://devfeed.tech/tags/embedded.md>), [embedded-systems](<https://devfeed.tech/tags/embedded-systems.md>), [event](<https://devfeed.tech/tags/event.md>), [events](<https://devfeed.tech/tags/events.md>), [meetup](<https://devfeed.tech/tags/meetup.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [open-source-community](<https://devfeed.tech/tags/open-source-community.md>), [presentations](<https://devfeed.tech/tags/presentations.md>), [technical](<https://devfeed.tech/tags/technical.md>), [university](<https://devfeed.tech/tags/university.md>), [workshop](<https://devfeed.tech/tags/workshop.md>), [zephyr](<https://devfeed.tech/tags/zephyr.md>)

### AI overview

The Hagenberg Zephyr Project meetup combines a beginner hands-on workshop with evening technical talks on November 18, 2026, at the University of Applied Sciences Upper Austria. The 3-5 PM workshop introduces Zephyr RTOS through a blinking LED example in native_sim and on hardware, shell exercises, and a connectivity example. Participants must bring a computer and complete the Getting Started Guide beforehand, which requires downloading several GB of data. The 5:30-8 PM meetup features talks on rapid prototyping and TinyML, reviving a toy robot with a Bluetooth LE remote, and booting Zephyr on an FPGA with an open-source RISC-V CPU, including a live demo. Registration is required and seats are limited.

### Source excerpt

The Zephyr Project community is coming together for an in-person meetup in Hagenberg, Austria! Join us on Wednesday, November 18, 2026, from 3:00-8:00 PM CET at the University of Applied Sciences Upper Austria, Campus Hagenberg. The event will feature a beginner-friendly hands-on workshop, technical presentations, and opportunities to connect with the local embedded systems and open source community.

## What to Expect at the Zephyr Project Meetup (November 18, 2026) - Hagenberg, Austria

DevFeed: [What to Expect at the Zephyr Project Meetup (November 18, 2026) - Hagenberg, Austria](<https://devfeed.tech/articles/what-to-expect-at-the-zephyr-project-meetup-november-18-2026-hagenberg-austria-136405.md>)

Original publisher: [Read original article](<https://www.zephyrproject.org/what-to-expect-at-the-zephyr-project-meetup-november-18-2026-hagenberg-austria/>)

Author: Susan Remmert

Published: 2026-11-16T21:25:01Z

Content type: news

Language: en

Sources: [Zephyr Project](<https://devfeed.tech/sources/zephyr-project.md>)

Topics: [Zephyr RTOS](<https://devfeed.tech/topics/zephyr-rtos.md>), [tinyml](<https://devfeed.tech/topics/tinyml.md>), [Open Source](<https://devfeed.tech/topics/open-source.md>), [developer-relations](<https://devfeed.tech/topics/developer-relations.md>), [Machine learning](<https://devfeed.tech/topics/machine-learning.md>)

Tags: [2026](<https://devfeed.tech/tags/2026.md>), [austria](<https://devfeed.tech/tags/austria.md>), [beginner](<https://devfeed.tech/tags/beginner.md>), [blog](<https://devfeed.tech/tags/blog.md>), [bluetooth](<https://devfeed.tech/tags/bluetooth.md>), [bluetooth-le](<https://devfeed.tech/tags/bluetooth-le.md>), [board](<https://devfeed.tech/tags/board.md>), [boot](<https://devfeed.tech/tags/boot.md>), [community](<https://devfeed.tech/tags/community.md>), [connectivity](<https://devfeed.tech/tags/connectivity.md>), [contributors](<https://devfeed.tech/tags/contributors.md>), [embedded](<https://devfeed.tech/tags/embedded.md>), [embedded-systems](<https://devfeed.tech/tags/embedded-systems.md>), [events](<https://devfeed.tech/tags/events.md>), [meetup](<https://devfeed.tech/tags/meetup.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [open-source-community](<https://devfeed.tech/tags/open-source-community.md>), [presentations](<https://devfeed.tech/tags/presentations.md>), [technical](<https://devfeed.tech/tags/technical.md>), [university](<https://devfeed.tech/tags/university.md>), [workshop](<https://devfeed.tech/tags/workshop.md>), [zephyr](<https://devfeed.tech/tags/zephyr.md>)

### AI overview

The Zephyr Project meetup in Hagenberg on November 18, 2026 combines a beginner hands-on workshop with evening technical talks. The workshop (3:00-5:00 PM CET) covers a blinking LED on native_sim and hardware, Zephyr's shell, and a connectivity example; attendees need a computer and should complete the Getting Started Guide beforehand, which requires downloading several GB. The meetup (5:30-8:00 PM) features talks on rapid prototyping and TinyML, reviving a toy robot as a Bluetooth LE remote, and running Zephyr with an open-source RISC-V CPU on an FPGA, including boot flow, driver/device-tree integration, and a live demo. Registration is required and seats are limited.

### Source excerpt

The Zephyr Project community is coming together for an in-person meetup in Hagenberg, Austria! Join us on Wednesday, November 18, 2026, from 3:00-8:00 PM CET at the University of Applied Sciences Upper Austria, Campus Hagenberg. The event will feature a beginner-friendly hands-on workshop, technical presentations, and opportunities to connect with the local embedded systems and open source community.

## The Sound of Smarter Farming

DevFeed: [The Sound of Smarter Farming](<https://devfeed.tech/articles/the-sound-of-smarter-farming-135951.md>)

Original publisher: [Read original article](<https://www.eetimes.com/the-sound-of-smarter-farming/>)

Author: Rebecca Pool

Published: 2026-11-16T06:59:20Z

Content type: news

Language: en

Sources: [EE Times](<https://devfeed.tech/sources/ee-times.md>)

Topics: [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [Physical AI](<https://devfeed.tech/topics/physical-ai.md>), [tinyml](<https://devfeed.tech/topics/tinyml.md>)

Tags: [acoustic-sensing](<https://devfeed.tech/tags/acoustic-sensing.md>), [ai](<https://devfeed.tech/tags/ai.md>), [analog](<https://devfeed.tech/tags/analog.md>), [analysis](<https://devfeed.tech/tags/analysis.md>), [business](<https://devfeed.tech/tags/business.md>), [community](<https://devfeed.tech/tags/community.md>), [designline](<https://devfeed.tech/tags/designline.md>), [digital](<https://devfeed.tech/tags/digital.md>), [education](<https://devfeed.tech/tags/education.md>), [ee-times](<https://devfeed.tech/tags/ee-times.md>), [ee-times-university](<https://devfeed.tech/tags/ee-times-university.md>), [eelife](<https://devfeed.tech/tags/eelife.md>), [eelive-electronic-design](<https://devfeed.tech/tags/eelive-electronic-design.md>), [eet](<https://devfeed.tech/tags/eet.md>), [electrical-engineering](<https://devfeed.tech/tags/electrical-engineering.md>), [electronic-components](<https://devfeed.tech/tags/electronic-components.md>), [electronic-engineering-times](<https://devfeed.tech/tags/electronic-engineering-times.md>), [electronics](<https://devfeed.tech/tags/electronics.md>), [embedded-systems](<https://devfeed.tech/tags/embedded-systems.md>), [farming](<https://devfeed.tech/tags/farming.md>), [fundamentals-courses](<https://devfeed.tech/tags/fundamentals-courses.md>), [learning](<https://devfeed.tech/tags/learning.md>), [mems-sensors](<https://devfeed.tech/tags/mems-sensors.md>), [microphones](<https://devfeed.tech/tags/microphones.md>), [mixed-signal](<https://devfeed.tech/tags/mixed-signal.md>), [news](<https://devfeed.tech/tags/news.md>), [opinion](<https://devfeed.tech/tags/opinion.md>), [part-search](<https://devfeed.tech/tags/part-search.md>), [plants](<https://devfeed.tech/tags/plants.md>), [product-search](<https://devfeed.tech/tags/product-search.md>), [semiconductors](<https://devfeed.tech/tags/semiconductors.md>), [sensor-processing](<https://devfeed.tech/tags/sensor-processing.md>), [sensors](<https://devfeed.tech/tags/sensors.md>), [smart-farming](<https://devfeed.tech/tags/smart-farming.md>), [software](<https://devfeed.tech/tags/software.md>), [sound](<https://devfeed.tech/tags/sound.md>), [tech-papers](<https://devfeed.tech/tags/tech-papers.md>), [technology](<https://devfeed.tech/tags/technology.md>), [threats](<https://devfeed.tech/tags/threats.md>), [video](<https://devfeed.tech/tags/video.md>)

### AI overview

Acoustic sensors and machine learning are being used to measure farmland biodiversity, pollinator activity, and crop pests through sound and vibration. Chirrup.ai records birdsong over two to three weeks and analyzes it remotely for annual biodiversity snapshots; this prioritizes battery life and cost, while species predictions are checked by ornithologists and ecologists. AgriSound's solar-powered Polly sensor analyzes bee wingbeats locally and sends pollinator data to the cloud hourly, providing faster results for time-sensitive flowering periods. BugBug clips piezoelectric sensors to plant stems to capture feeding vibrations; its custom ML uses unsupervised learning and contextual data because individual insect sounds are difficult to label. The systems illustrate different sampling and processing choices, and the excerpt describes pest early warning and treatment timing as goals still under investigation.

### Source excerpt

From birdsong and pollinator wingbeats to the vibrations of insects feeding inside plants, acoustic sensors and AI are helping farmers monitor biodiversity, improve pollination, and detect threats to crops. The post The Sound of Smarter Farming appeared first on EE Times.

## RL$^{\\mathrm{RT}}$: On-Device Reinforcement Learning Runtime

DevFeed: [RL$^{\\mathrm{RT}}$: On-Device Reinforcement Learning Runtime](<https://devfeed.tech/articles/rl-mathrm-rt-on-device-reinforcement-learning-runtime-135161.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.18035>)

Author: Zexin Li, Xinwei Gao, Arvind Easwaran

Published: 2026-11-16T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>), [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Reinforcement learning](<https://devfeed.tech/topics/reinforcement-learning.md>), [On-device AI](<https://devfeed.tech/topics/on-device-ai.md>), [tinyml](<https://devfeed.tech/topics/tinyml.md>), [Machine learning](<https://devfeed.tech/topics/machine-learning.md>)

Tags: [actor](<https://devfeed.tech/tags/actor.md>), [admission](<https://devfeed.tech/tags/admission.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [embedded](<https://devfeed.tech/tags/embedded.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [jetson](<https://devfeed.tech/tags/jetson.md>), [learner](<https://devfeed.tech/tags/learner.md>), [on-device](<https://devfeed.tech/tags/on-device.md>), [refresh](<https://devfeed.tech/tags/refresh.md>), [reinforcement-learning](<https://devfeed.tech/tags/reinforcement-learning.md>), [runtime](<https://devfeed.tech/tags/runtime.md>), [scheduler](<https://devfeed.tech/tags/scheduler.md>), [training](<https://devfeed.tech/tags/training.md>)

### AI overview

RL$^{\mathrm{RT}}$ coordinates a periodic actor and learner sharing an embedded GPU by making their coupling tunable. A refresh bound limits actor releases between weight updates, while a GPU permit with slack-aware admission prioritizes actor work and uses spare time for training on top of a base scheduler. On two Jetson boards, accounting for the real refresh cost, it kept deadline misses for a 16.8-million-parameter policy below 1%, compared with 85-100% without the runtime.

### Source excerpt

arXiv:2611.18035v1 Announce Type: cross Abstract: On-device reinforcement learning runs a periodic actor and a learner on one embedded GPU shared with other workloads. Because the learner produces the actor's weights, existing designs stall the actor or let its policy drift. We present RL$^{\mathrm{RT}}$, an on-device runtime that makes this coupling a tunable parameter: a refresh bound limits actor releases between refreshes, and a GPU permit with slack-aware admission prioritises the actor and fills spare time with training atop any base scheduler. Under the real refresh cost on two Jetson boards, RL$^{\mathrm{RT}}$ keeps a 16.8 M-parameter policy's deadline misses below 1%, against 85 to 100% without it.

## Designing Edge AI SoCs At Near-Threshold: Why Foundation IP Can Help You Break Down The Power Wall

DevFeed: [Designing Edge AI SoCs At Near-Threshold: Why Foundation IP Can Help You Break Down The Power Wall](<https://devfeed.tech/articles/designing-edge-ai-socs-at-near-threshold-why-foundation-ip-can-help-you-break-down-the-power-wall-130164.md>)

Original publisher: [Read original article](<https://semiengineering.com/designing-edge-ai-socs-at-near-threshold-why-foundation-ip-can-help-you-break-down-the-power-wall/>)

Author: Andrew Appleby

Published: 2026-11-12T08:05:04Z

Content type: article

Language: en

Sources: [Semiconductor Engineering](<https://devfeed.tech/sources/semiconductor-engineering.md>)

Topics: [tinyml](<https://devfeed.tech/topics/tinyml.md>), [Sustainable Computing](<https://devfeed.tech/topics/sustainable-computing.md>)

Tags: [3nm](<https://devfeed.tech/tags/3nm.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-networking](<https://devfeed.tech/tags/ai-networking.md>), [ai-workloads](<https://devfeed.tech/tags/ai-workloads.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [area](<https://devfeed.tech/tags/area.md>), [bandwidth](<https://devfeed.tech/tags/bandwidth.md>), [behavior](<https://devfeed.tech/tags/behavior.md>), [benefit](<https://devfeed.tech/tags/benefit.md>), [blocks](<https://devfeed.tech/tags/blocks.md>), [blogs-lphp](<https://devfeed.tech/tags/blogs-lphp.md>), [edge-ai](<https://devfeed.tech/tags/edge-ai.md>), [energy](<https://devfeed.tech/tags/energy.md>), [foundation-ip](<https://devfeed.tech/tags/foundation-ip.md>), [ip](<https://devfeed.tech/tags/ip.md>), [ip-and-lp-in-socs](<https://devfeed.tech/tags/ip-and-lp-in-socs.md>), [logic-libraries](<https://devfeed.tech/tags/logic-libraries.md>), [low-power-high-performance](<https://devfeed.tech/tags/low-power-high-performance.md>), [low-voltage-design](<https://devfeed.tech/tags/low-voltage-design.md>), [near-threshold](<https://devfeed.tech/tags/near-threshold.md>), [performance](<https://devfeed.tech/tags/performance.md>), [power](<https://devfeed.tech/tags/power.md>), [real-time](<https://devfeed.tech/tags/real-time.md>), [scaling](<https://devfeed.tech/tags/scaling.md>), [synopsys](<https://devfeed.tech/tags/synopsys.md>), [voltage](<https://devfeed.tech/tags/voltage.md>), [voltage-scaling](<https://devfeed.tech/tags/voltage-scaling.md>)

### AI overview

Near-threshold operation around 0.4-0.5 V can improve energy per operation for edge AI SoCs, but delay and variability rise and noise margins shrink, making predictable design a cross-stack challenge. The excerpt says success depends on carefully selected and silicon-validated Foundation IP, accurate modeling, and close alignment with EDA flows. Memory can set system Vmin; a dual-rail architecture is cited as decoupling logic and memory voltage, with a 38 Mb/mm² SRAM validated at 380/540 mV in 3nm FinFET. It also describes different memory needs across MAC tiles, activation units, and shared buffers, and notes that I/O reliability can constrain system Vmin. In one high-volume AI networking SoC engagement, tuned low-voltage logic cells, custom level shifters, a dual-rail memory compiler, selected PVT corners, and application engineering support were used; the stated result was predictable execution without schedule or silicon risk. No quantitative power or performance result is provided.

### Source excerpt

Traditional voltage scaling is no longer sufficient to deliver real-time performance within tight energy and thermal limits. The post Designing Edge AI SoCs At Near-Threshold: Why Foundation IP Can Help You Break Down The Power Wall appeared first on Semiconductor Engineering.

## People Detection in Confined Spaces: Elevator Application

DevFeed: [People Detection in Confined Spaces: Elevator Application](<https://devfeed.tech/articles/people-detection-in-confined-spaces-elevator-application-127081.md>)

Original publisher: [Read original article](<https://www.eetimes.com/people-detection-in-confined-spaces-elevator-application/>)

Author: Thomas Parton

Published: 2026-11-10T15:00:00Z

Content type: article

Language: en

Sources: [EE Times](<https://devfeed.tech/sources/ee-times.md>)

Topics: [Embedded Systems](<https://devfeed.tech/topics/embedded-systems.md>), [Library](<https://devfeed.tech/topics/library.md>), [tinyml](<https://devfeed.tech/topics/tinyml.md>)

Tags: [analog](<https://devfeed.tech/tags/analog.md>), [analysis](<https://devfeed.tech/tags/analysis.md>), [application](<https://devfeed.tech/tags/application.md>), [array](<https://devfeed.tech/tags/array.md>), [business](<https://devfeed.tech/tags/business.md>), [cabins](<https://devfeed.tech/tags/cabins.md>), [cameras](<https://devfeed.tech/tags/cameras.md>), [cells](<https://devfeed.tech/tags/cells.md>), [cloud](<https://devfeed.tech/tags/cloud.md>), [community](<https://devfeed.tech/tags/community.md>), [design](<https://devfeed.tech/tags/design.md>), [designline](<https://devfeed.tech/tags/designline.md>), [detection](<https://devfeed.tech/tags/detection.md>), [digital](<https://devfeed.tech/tags/digital.md>), [edge](<https://devfeed.tech/tags/edge.md>), [education](<https://devfeed.tech/tags/education.md>), [ee-times](<https://devfeed.tech/tags/ee-times.md>), [ee-times-university](<https://devfeed.tech/tags/ee-times-university.md>), [eelife](<https://devfeed.tech/tags/eelife.md>), [eelive-electronic-design](<https://devfeed.tech/tags/eelive-electronic-design.md>), [eet](<https://devfeed.tech/tags/eet.md>), [electrical-engineering](<https://devfeed.tech/tags/electrical-engineering.md>), [electronic-components](<https://devfeed.tech/tags/electronic-components.md>), [electronic-engineering-times](<https://devfeed.tech/tags/electronic-engineering-times.md>), [electronics](<https://devfeed.tech/tags/electronics.md>), [elevator-automation](<https://devfeed.tech/tags/elevator-automation.md>), [embedded](<https://devfeed.tech/tags/embedded.md>), [embedded-systems](<https://devfeed.tech/tags/embedded-systems.md>), [far-infrared-fir-thermal-sensor](<https://devfeed.tech/tags/far-infrared-fir-thermal-sensor.md>), [fundamentals-courses](<https://devfeed.tech/tags/fundamentals-courses.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [imaging](<https://devfeed.tech/tags/imaging.md>), [learning](<https://devfeed.tech/tags/learning.md>), [library](<https://devfeed.tech/tags/library.md>), [load](<https://devfeed.tech/tags/load.md>), [local](<https://devfeed.tech/tags/local.md>), [mixed-signal](<https://devfeed.tech/tags/mixed-signal.md>), [mlx90642](<https://devfeed.tech/tags/mlx90642.md>), [monitoring](<https://devfeed.tech/tags/monitoring.md>), [news](<https://devfeed.tech/tags/news.md>), [opinion](<https://devfeed.tech/tags/opinion.md>), [part-search](<https://devfeed.tech/tags/part-search.md>), [people](<https://devfeed.tech/tags/people.md>), [people-detection-and-counting](<https://devfeed.tech/tags/people-detection-and-counting.md>), [privacy-by-design](<https://devfeed.tech/tags/privacy-by-design.md>), [product-search](<https://devfeed.tech/tags/product-search.md>), [semiconductors](<https://devfeed.tech/tags/semiconductors.md>), [software](<https://devfeed.tech/tags/software.md>), [spaces](<https://devfeed.tech/tags/spaces.md>), [tech-papers](<https://devfeed.tech/tags/tech-papers.md>), [technology](<https://devfeed.tech/tags/technology.md>), [video](<https://devfeed.tech/tags/video.md>), [white-papers](<https://devfeed.tech/tags/white-papers.md>)

### AI overview

The paper describes an elevator occupancy-monitoring system pairing Melexis's MLX90642 far-infrared thermal array with a hardware-agnostic People Detection Library. It processes thermal data locally, without cloud streaming, using dynamic background suppression, thermal blob separation and spatial tracking. It evaluates Simple Mode for ultra-low-power Cortex-M0+/M4 MCUs and Default (Full) Mode for advanced multi-passenger counting on mainstream embedded processors. It also provides prototyping and evaluation guidance for the DVK90642 kit. The excerpt does not provide measured performance results or details of the alternatives' limitations.

### Source excerpt

This technical white paper details an embedded occupancy-monitoring solution for elevator cabins combining the Melexis MLX90642 Far-Infrared (FIR) thermal sensor array with a hardware-agnostic People Detection Library. Engineered to overcome the limitations of optical cameras, PIR sensors, and floor load cells, the solution provides guaranteed privacy-by-design via local edge processing and thermal imaging without cloud [...] The post People Detection in Confined Spaces: Elevator Application appeared first on EE Times.

## Neural Activity Synchronization-Based Fast Local Learnable Spiking Neural Network Architecture

DevFeed: [Neural Activity Synchronization-Based Fast Local Learnable Spiking Neural Network Architecture](<https://devfeed.tech/articles/neural-activity-synchronization-based-fast-local-learnable-spiking-neural-network-architecture-126065.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.11053>)

Author: Pranith Dutta, Sk Wasim Rajjak, Siddhanta Ray, Santlal Prajapati, Manobendra Nath Mondal

Published: 2026-11-10T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Neural Network](<https://devfeed.tech/topics/neural-network.md>), [Deep neural networks](<https://devfeed.tech/topics/deep-neural-networks.md>), [Deep learning](<https://devfeed.tech/topics/deep-learning.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [tinyml](<https://devfeed.tech/topics/tinyml.md>)

Tags: [architecture](<https://devfeed.tech/tags/architecture.md>), [artificial-neural-networks](<https://devfeed.tech/tags/artificial-neural-networks.md>), [backpropagation](<https://devfeed.tech/tags/backpropagation.md>), [battery](<https://devfeed.tech/tags/battery.md>), [classification](<https://devfeed.tech/tags/classification.md>), [classification-accuracy](<https://devfeed.tech/tags/classification-accuracy.md>), [computational-complexity](<https://devfeed.tech/tags/computational-complexity.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-ne](<https://devfeed.tech/tags/cs-ne.md>), [dataset](<https://devfeed.tech/tags/dataset.md>), [devices](<https://devfeed.tech/tags/devices.md>), [edge](<https://devfeed.tech/tags/edge.md>), [edge-devices](<https://devfeed.tech/tags/edge-devices.md>), [energy](<https://devfeed.tech/tags/energy.md>), [energy-consumption](<https://devfeed.tech/tags/energy-consumption.md>), [fast](<https://devfeed.tech/tags/fast.md>), [framework](<https://devfeed.tech/tags/framework.md>), [local](<https://devfeed.tech/tags/local.md>), [network](<https://devfeed.tech/tags/network.md>), [neural](<https://devfeed.tech/tags/neural.md>), [neural-network](<https://devfeed.tech/tags/neural-network.md>), [synchronization](<https://devfeed.tech/tags/synchronization.md>)

### AI overview

The proposed Neural Activity Synchronization (NAS)-inspired framework trains multilayer spiking neural networks with locally computed gradients: each layer updates its own weights without backpropagating gradients through the network. The authors present this as a way to reduce training complexity, latency, and energy use compared with the global, computationally expensive BPTT approach. On MNIST, the model achieves 97.78% classification accuracy; the abstract also reports robust performance across Gaussian and salt-and-pepper noise levels, without specifying the levels or detailed results.

### Source excerpt

arXiv:2611.11053v1 Announce Type: cross Abstract: Implementing intelligent learning systems on low- power edge devices is challenging due to the high compu- tational complexity, training latency, and energy demands of traditional Artificial Neural Networks (ANNs). Spiking Neural Networks (SNNs), biologically plausible and spike-driven, come as an alternative. However, SNN training commonly relies on backpropagation through time (BPTT), which is computationally expensive, inherently global, and unsuitable for battery-driven devices. This work proposes a novel local learning framework for multi-layered SNNs inspired by Neural Activity Synchronization (NAS). Each layer updates its own weights using locally com- puted gradients. No gradient backpropagation is needed. This significantly reduces computational complexity, training latency, and energy consumption. The proposed SNN model is trained and evaluated on the MNIST dataset, achieving a classification accuracy of 97.78%. The model is also analyzed with Gaussian and salt-and-pepper noise, demonstrating robust performance across different noise levels.

## Sensitivity-Gated Accumulator Caching for On-Device Learning of Low-Bit Hand Networks

DevFeed: [Sensitivity-Gated Accumulator Caching for On-Device Learning of Low-Bit Hand Networks](<https://devfeed.tech/articles/sensitivity-gated-accumulator-caching-for-on-device-learning-of-low-bit-hand-networks-125294.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.10991>)

Author: Soroor Ghandali, Kanika Mamgain, Manjot Kaur, Akanksha Shukla, Shashira Guntuka

Published: 2026-11-10T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [quantization](<https://devfeed.tech/topics/quantization.md>), [On-device AI](<https://devfeed.tech/topics/on-device-ai.md>), [tinyml](<https://devfeed.tech/topics/tinyml.md>)

Tags: [caching](<https://devfeed.tech/tags/caching.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [learning](<https://devfeed.tech/tags/learning.md>), [networks](<https://devfeed.tech/tags/networks.md>), [on-device](<https://devfeed.tech/tags/on-device.md>), [on-device-learning](<https://devfeed.tech/tags/on-device-learning.md>), [quantization](<https://devfeed.tech/tags/quantization.md>)

### AI overview

Sensitivity-gated accumulator caching (SGAC) adapts low-bit network weights while keeping residual state only for selected groups. It uses a cache of 8-bit accumulators to collect sub-step updates for groups ranked by Fisher-weighted squared code step, then writes whole steps into the codes; evicted accumulators are flushed by stochastic rounding, while groups without a cache line do not update. The authors report unbiased write-back and, under a local quadratic model, explain that rounding is worthwhile where expected gradients are large relative to curvature. On a 383k-parameter hand-keypoint network, SGAC uses 8.1 state bits per weight and achieves the highest mean accuracy among nine deployed low-bit networks on two simulated cameras. Across twelve ImageNet-C corruptions, it leads seven deployed methods in average clean and corrupted accuracy, with confidence intervals excluding zero, and matches a per-weight 8-bit accumulator using 56% as much state.

### Source excerpt

arXiv:2611.10991v1 Announce Type: cross Abstract: Small hand-tracking networks are deployed as 4- and 8-bit codes, and they should keep adapting to a user's camera after deployment, on the device and within roughly the memory of the model itself. Quantization-aware training cannot do this cheaply, because it keeps a 32-bit master copy of every weight together with its optimizer state, about 96 bits per weight. Training the codes directly with stochastic rounding needs no copy but injects rounding noise at every step, and the methods that remove this noise keep a residual buffer for every weight. We introduce sensitivity-gated accumulator caching (SGAC), which keeps residual state for only a small and changing set of weight groups. A set-associative cache of 8-bit accumulators collects the sub-step updates of the groups with the largest Fisher-weighted squared code step and writes them into the codes as whole steps; an evicted line is flushed by stochastic rounding, and a group without a line does not move until it is admitted. We prove that the write-back is unbiased and show, under a local quadratic model, that rounding a sub-step update costs loss of the same order as the update's own gain, so that rounding pays only where the expected gradient is large relative to the curvature. On a 383k-parameter hand-keypoint network, SGAC trains with 8.1 bits of state per weight. Scored on the 4/8-bit network that the device runs after adaptation, it has the highest mean accuracy of nine deployed low-bit networks on two simulated cameras. On twelve ImageNet-C corruptions it has the highest average of corrupted and clean accuracy of seven deployed methods, ahead of the PEQA-style, ECO-style and direct low-bit baselines with confidence intervals that exclude zero, and it matches an 8-bit accumulator kept for every weight with 56% of that state.

## ConvLUT: Rethinking Convolutional Architectures for Ultra-Low-Latency LUT-Based Inference

DevFeed: [ConvLUT: Rethinking Convolutional Architectures for Ultra-Low-Latency LUT-Based Inference](<https://devfeed.tech/articles/convlut-rethinking-convolutional-architectures-for-ultra-low-latency-lut-based-inference-125395.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.11507>)

Author: Marta Andronic, Olivia Weng, Danial Zuberi, Ryan Kastner, George A. Constantinides

Published: 2026-11-10T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [tinyml](<https://devfeed.tech/topics/tinyml.md>)

Tags: [architectures](<https://devfeed.tech/tags/architectures.md>), [audio](<https://devfeed.tech/tags/audio.md>), [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [cifar-10](<https://devfeed.tech/tags/cifar-10.md>), [convergence](<https://devfeed.tech/tags/convergence.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [fpgas](<https://devfeed.tech/tags/fpgas.md>), [framework](<https://devfeed.tech/tags/framework.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [inference](<https://devfeed.tech/tags/inference.md>), [keyword-spotting](<https://devfeed.tech/tags/keyword-spotting.md>), [latency](<https://devfeed.tech/tags/latency.md>), [layer](<https://devfeed.tech/tags/layer.md>), [logic](<https://devfeed.tech/tags/logic.md>), [lookup](<https://devfeed.tech/tags/lookup.md>), [low-latency](<https://devfeed.tech/tags/low-latency.md>), [low-latency-inference](<https://devfeed.tech/tags/low-latency-inference.md>), [lut](<https://devfeed.tech/tags/lut.md>), [mlp](<https://devfeed.tech/tags/mlp.md>), [mlperf](<https://devfeed.tech/tags/mlperf.md>), [model](<https://devfeed.tech/tags/model.md>)

### AI overview

ConvLUT extends LUT-based FPGA inference from MLP-style networks to convolutional vision and audio models. It offers standard convolution built around hierarchical LUT trees and a depthwise separable model following MobileNet's design. Its framework supports standard, depthwise, and pointwise convolutions with explicit fan-in constraints; additional techniques address large receptive fields, replace spatial pooling with channel-partitioned LUT pooling, and use hardware-aware iterative pruning. On CIFAR-10, Visual Wake Words, and Keyword Spotting, the fully unrolled models report inference latency below about 50 ns. On CIFAR-10, the authors report at least 8,844x lower latency than non-logic-based prior implementations and at least a 20 percentage-point test-accuracy advantage over prior MLP-style LUT models.

### Source excerpt

arXiv:2611.11507v1 Announce Type: new Abstract: Lookup-table (LUT)-based neural networks (NNs) offer a compelling path to ultra-low-latency inference on FPGAs. However, existing LUT-based neural networks have largely been restricted to MLP-style topologies, leaving structured vision and audio tasks underexplored. We present ConvLUT, a family of fully unrolled LUT-mappable convolutional architectures targeting the MLPerf Tiny benchmarks. It is comprised of a standard convolutional model built around a hierarchical tree of LUTs and a depthwise separable model following the MobileNet design philosophy. These architectures are enabled by four key contributions: a unified LUT-based convolutional framework supporting standard, depthwise, and pointwise convolutions under explicit fan-in constraints; a LUT-tree that addresses the $k^2c$ receptive field dimensionality bottleneck; a channel-partitioned LUT pooling layer that replaces conventional spatial pooling; and a hardware-aware iterative pruning strategy for stable convergence at scale. Evaluated on CIFAR-10, Visual Wake Words, and Keyword Spotting, ConvLUT achieves competitive accuracy with fully unrolled inference latencies below $\sim 50$ ns. On CIFAR-10 it reaches at least a $8844\times$ latency reduction over non-logic-based prior implementations, and outperforms existing MLP-style LUT-based models on the same task by at least $20$ pp in test accuracy.

## AI Needs a Body: Three Builds from Tomorrow Lab® Where AI and Hardware Work as One

DevFeed: [AI Needs a Body: Three Builds from Tomorrow Lab® Where AI and Hardware Work as One](<https://devfeed.tech/articles/ai-needs-a-body-three-builds-from-tomorrow-lab-where-ai-and-hardware-work-as-one-123855.md>)

Original publisher: [Read original article](<https://hackernoon.com/ai-needs-a-body-three-builds-from-tomorrow-labr-where-ai-and-hardware-work-as-one?source=rss>)

Author: Jon Stojan Journalist

Published: 2026-11-09T11:30:35Z

Content type: article

Language: en

Sources: [HackerNoon](<https://devfeed.tech/sources/hackernoon.md>)

Topics: [Physical AI](<https://devfeed.tech/topics/physical-ai.md>), [Embedded Systems](<https://devfeed.tech/topics/embedded-systems.md>), [Computer vision](<https://devfeed.tech/topics/computer-vision.md>), [tinyml](<https://devfeed.tech/topics/tinyml.md>), [UNO Q](<https://devfeed.tech/topics/uno-q.md>), [IoT predictive analytics](<https://devfeed.tech/topics/iot-predictive-analytics.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-hardware-integration](<https://devfeed.tech/tags/ai-hardware-integration.md>), [ai-model](<https://devfeed.tech/tags/ai-model.md>), [ai-powered-iot-devices](<https://devfeed.tech/tags/ai-powered-iot-devices.md>), [ai-products](<https://devfeed.tech/tags/ai-products.md>), [ai-vision](<https://devfeed.tech/tags/ai-vision.md>), [algorithms](<https://devfeed.tech/tags/algorithms.md>), [analog-devices](<https://devfeed.tech/tags/analog-devices.md>), [app-lab](<https://devfeed.tech/tags/app-lab.md>), [arduino](<https://devfeed.tech/tags/arduino.md>), [arduino-app-lab](<https://devfeed.tech/tags/arduino-app-lab.md>), [arduino-uno-q](<https://devfeed.tech/tags/arduino-uno-q.md>), [arm](<https://devfeed.tech/tags/arm.md>), [cloud](<https://devfeed.tech/tags/cloud.md>), [cortex-m4](<https://devfeed.tech/tags/cortex-m4.md>), [detection](<https://devfeed.tech/tags/detection.md>), [edge](<https://devfeed.tech/tags/edge.md>), [edge-ai-hardware](<https://devfeed.tech/tags/edge-ai-hardware.md>), [embedded-computer-vision](<https://devfeed.tech/tags/embedded-computer-vision.md>), [golioth-iot-platform](<https://devfeed.tech/tags/golioth-iot-platform.md>), [good-company](<https://devfeed.tech/tags/good-company.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [intelligence](<https://devfeed.tech/tags/intelligence.md>), [leak](<https://devfeed.tech/tags/leak.md>), [low-power-machine-learning](<https://devfeed.tech/tags/low-power-machine-learning.md>), [physical-ai-applications](<https://devfeed.tech/tags/physical-ai-applications.md>), [recognition](<https://devfeed.tech/tags/recognition.md>), [vision](<https://devfeed.tech/tags/vision.md>)

### AI overview

Three Tomorrow Lab builds pair physical problems with different AI hardware and platforms. A drawbridge watcher uses a webcam and Arduino UNO Q: Debian Linux handles the vision application while its STM32U585 microcontroller supports real-time control. A pet feeder uses the Analog Devices MAX78000 to recognize pets and decide locally, offline, at low power. A leak-detection network uses Golioth for centralized device management, sensor data streaming, pattern analysis, and remote firmware and detection-logic updates. The article's central design point is to match sensing, compute, control, and connectivity to each task from the start.

### Source excerpt

Explore three Tomorrow Lab projects combining AI and hardware, from drawbridge vision and pet recognition to connected leak detection and edge intelligence.

## Embedded Arena: Iterative Optimization via Hardware Feedback

DevFeed: [Embedded Arena: Iterative Optimization via Hardware Feedback](<https://devfeed.tech/articles/embedded-arena-iterative-optimization-via-hardware-feedback-124442.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2606.16190>)

Author: Zhihan Zhang, Alexander Le Metzger, Jiuyang Lyu, Chun-Cheng Chang, Jiayi Shao, Yujia Liu, Emmanuel Azuh Mensah, Edward Wang, Kurtis Heimerl, Gregory D. Abowd, Shwetak Patel, Natasha Jaques, Vikram Iye

Published: 2026-11-09T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [tinyml](<https://devfeed.tech/topics/tinyml.md>), [Embedded Systems](<https://devfeed.tech/topics/embedded-systems.md>), [Embedded Software Dev](<https://devfeed.tech/topics/embedded-software-dev.md>), [Multi-step AI processes](<https://devfeed.tech/topics/multi-step-ai-processes.md>)

Tags: [agent](<https://devfeed.tech/tags/agent.md>), [ai](<https://devfeed.tech/tags/ai.md>), [arena](<https://devfeed.tech/tags/arena.md>), [audio](<https://devfeed.tech/tags/audio.md>), [battery](<https://devfeed.tech/tags/battery.md>), [camera](<https://devfeed.tech/tags/camera.md>), [compression](<https://devfeed.tech/tags/compression.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-ar](<https://devfeed.tech/tags/cs-ar.md>), [deployment](<https://devfeed.tech/tags/deployment.md>), [detection](<https://devfeed.tech/tags/detection.md>), [development](<https://devfeed.tech/tags/development.md>), [edge](<https://devfeed.tech/tags/edge.md>), [elk](<https://devfeed.tech/tags/elk.md>), [embedded](<https://devfeed.tech/tags/embedded.md>), [error](<https://devfeed.tech/tags/error.md>), [experts](<https://devfeed.tech/tags/experts.md>), [feedback](<https://devfeed.tech/tags/feedback.md>), [firmware](<https://devfeed.tech/tags/firmware.md>), [flashing](<https://devfeed.tech/tags/flashing.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [optimization](<https://devfeed.tech/tags/optimization.md>)

### AI overview

Embedded Arena uses a hardware-in-the-loop agent to optimize models and firmware for microcontrollers through repeated compilation, flashing, and measurement on real devices. The abstract reports that Claude Opus 4.7 and Gemini 3.1 Pro achieved 0% deployment success without hardware feedback; with feedback, the approach reached a first deployment within three iterations and could exceed expert results within seven. It reports 250x vision-model compression with under 3.3% accuracy loss and 400x audio compression with under 6% Feature Error Rate loss, including battery-free operation using solar harvesting. Demonstrations include an elk-detection camera trap at 96.7% accuracy and a phonetic-transcription wearable at 8.37% FER.

### Source excerpt

arXiv:2606.16190v2 Announce Type: replace-cross Abstract: As AI moves into the physical world, from robots to environmental sensors to medical wearables, it increasingly relies on local inference to meet latency, communication, or privacy constraints. Optimizing models for heterogeneous edge hardware such as microcontrollers (MCUs) requires simultaneously satisfying hard physical constraints on memory, power, and temperature while preserving accuracy, a multidimensional optimization that is today performed manually by experts. We ask whether an LLM agent can autonomously navigate this complex, multi-turn pipeline guided by real hardware feedback, and introduce a hardware-in-the-loop agent arena in which the agent iteratively refines both model and firmware by compiling, flashing, and measuring on real hardware, enabling closed-loop optimization. Frontier models, including Claude Opus 4.7 and Gemini 3.1 Pro, fail entirely without hardware feedback (0% deployment success), whereas our hardware-in-the-loop formulation achieves the first successful deployment within three iterations and can surpass human expert results within seven. This agentic co-optimization achieves 250x compression for vision models with <3.3% accuracy loss and 400x for audio with <6% Feature Error Rate loss, enabling battery-free operation on a commercial MCU via solar harvesting. We demonstrate practical impact in two real-world systems: an elk-detection camera trap (96.7% accuracy) and a phonetic-transcription wearable (8.37% FER) for child development research.

## 520 MHz STM32H7P4/STM32H7P5 Arm Cortex-M7 MCUs feature Hardware Signal Processor (HSP), optional Neural-ART NPU

DevFeed: [520 MHz STM32H7P4/STM32H7P5 Arm Cortex-M7 MCUs feature Hardware Signal Processor (HSP), optional Neural-ART NPU](<https://devfeed.tech/articles/520-mhz-stm32h7p4-stm32h7p5-arm-cortex-m7-mcus-feature-hardware-signal-processor-hsp-optional-neural-art-npu-121109.md>)

Original publisher: [Read original article](<https://www.cnx-software.com/2026/11/06/520-mhz-stm32h7p4-stm32h7p5-arm-cortex-m7-mcus-feature-hardware-signal-processor-hsp-optional-neural-art-npu/>)

Author: Debashis Das

Published: 2026-11-06T11:04:55Z

Content type: news

Language: en

Sources: [CNX Software - Embedded Systems News](<https://devfeed.tech/sources/cnx-software-embedded-systems-news.md>)

Topics: [Embedded Systems](<https://devfeed.tech/topics/embedded-systems.md>), [cpu](<https://devfeed.tech/topics/cpu.md>), [tinyml](<https://devfeed.tech/topics/tinyml.md>), [fourier analysis](<https://devfeed.tech/topics/fourier-analysis.md>)

Tags: [accelerators](<https://devfeed.tech/tags/accelerators.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-inference](<https://devfeed.tech/tags/ai-inference.md>), [arm](<https://devfeed.tech/tags/arm.md>), [can-fd](<https://devfeed.tech/tags/can-fd.md>), [cortex-m7](<https://devfeed.tech/tags/cortex-m7.md>), [development-board](<https://devfeed.tech/tags/development-board.md>), [dsp](<https://devfeed.tech/tags/dsp.md>), [edge-ai](<https://devfeed.tech/tags/edge-ai.md>), [flash](<https://devfeed.tech/tags/flash.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [hsp](<https://devfeed.tech/tags/hsp.md>), [industrial](<https://devfeed.tech/tags/industrial.md>), [mcu](<https://devfeed.tech/tags/mcu.md>), [mcus](<https://devfeed.tech/tags/mcus.md>), [motor-control](<https://devfeed.tech/tags/motor-control.md>), [npu](<https://devfeed.tech/tags/npu.md>), [optional](<https://devfeed.tech/tags/optional.md>), [power](<https://devfeed.tech/tags/power.md>), [pwm](<https://devfeed.tech/tags/pwm.md>), [sensing](<https://devfeed.tech/tags/sensing.md>), [sensors](<https://devfeed.tech/tags/sensors.md>), [signal-processing](<https://devfeed.tech/tags/signal-processing.md>), [sram](<https://devfeed.tech/tags/sram.md>), [stm32](<https://devfeed.tech/tags/stm32.md>), [stmicro](<https://devfeed.tech/tags/stmicro.md>), [stmicro-stm32](<https://devfeed.tech/tags/stmicro-stm32.md>), [tinyml](<https://devfeed.tech/tags/tinyml.md>), [video](<https://devfeed.tech/tags/video.md>)

### AI overview

The STM32H7P4 and STM32H7P5 are 520 MHz Cortex-M7 MCUs aimed at digital power, motor control, industrial sensing, and edge AI. Both include an HSP for matrix, vector, FFT/IFFT, and filtering work; ST says it runs FFT operations up to five times faster than Cortex-M7 software. The H7P5 adds a 520 MHz Neural-ART NPU rated at 85 GOPS, and ST says its AFCI model runs up to 40 times faster than a software implementation on the Cortex-M7. Both offer up to 2 MB flash, 736 KB SRAM, high-resolution timers with up to 24 PWM outputs, four 12-bit ADCs, and broad connectivity. Software includes STM32CubeH7P and motor-control and digital-power packages; two development boards were also announced. The MCUs are sampling for OEMs, with mass production planned for January 2027. Listed starting price is $4.43 for a 1 MB LQFP100 device at 10,000 units.

### Source excerpt

STMicroelectronics has introduced the STM32H7P4 and STM32H7P5, two 520 MHz Arm Cortex-M7 MCUs part of STM32H7P family designed for digital power, motor control, industrial sensing, and edge AI applications. Both feature a Hardware Signal Processor (HSP) for accelerating mathematical and signal-processing operations, while the STM32H7P5 also includes an 85 GOPS Neural-ART NPU for Edge AI inference/TinyML. The MCUs feature up to 2MB dual-bank flash, 736KB SRAM, two high-resolution timers with up to 24 PWM outputs, three advanced motor-control timers, and four 12-bit ADCs. Other features include CORDIC and FMAC mathematical accelerators, three CAN FD interfaces, USB OTG Full-Speed, and support for external memory through Octo-SPI and FMC interfaces. STM32H7P4 and STM32H7P5 specifications: MCU Core 32-bit Arm Cortex-M7 CPU @ up to 520 MHz with double-precision FPU, DSP instructions, and MPU 32 KB instruction cache and 32 KB data cache Performance - 2773 CoreMark (5.33 CoreMark/MHz) Mathematical coprocessors Hardware Signal Processor [...] The post 520 MHz STM32H7P4/STM32H7P5 Arm Cortex-M7 MCUs feature Hardware Signal Processor (HSP), optional Neural-ART NPU appeared first on CNX Software - Embedded Systems News.

## What to Expect at the Zephyr Project Meetup (November 13, 2026) - Munich, Germany

DevFeed: [What to Expect at the Zephyr Project Meetup (November 13, 2026) - Munich, Germany](<https://devfeed.tech/articles/what-to-expect-at-the-zephyr-project-meetup-november-13-2026-munich-germany-121808.md>)

Original publisher: [Read original article](<https://www.zephyrproject.org/what-to-expect-at-the-zephyr-project-meetup-november-13-2026-munich-germany/>)

Author: Susan Remmert

Published: 2026-11-06T05:00:04Z

Content type: news

Language: en

Sources: [Zephyr Project](<https://devfeed.tech/sources/zephyr-project.md>)

Topics: [Zephyr RTOS](<https://devfeed.tech/topics/zephyr-rtos.md>), [Open Source](<https://devfeed.tech/topics/open-source.md>), [tinyml](<https://devfeed.tech/topics/tinyml.md>), [cyber resilience act](<https://devfeed.tech/topics/cyber-resilience-act.md>)

Tags: [2026](<https://devfeed.tech/tags/2026.md>), [blog](<https://devfeed.tech/tags/blog.md>), [community](<https://devfeed.tech/tags/community.md>), [contributors](<https://devfeed.tech/tags/contributors.md>), [developers](<https://devfeed.tech/tags/developers.md>), [development](<https://devfeed.tech/tags/development.md>), [edge](<https://devfeed.tech/tags/edge.md>), [embedded](<https://devfeed.tech/tags/embedded.md>), [embedded-systems](<https://devfeed.tech/tags/embedded-systems.md>), [event](<https://devfeed.tech/tags/event.md>), [events](<https://devfeed.tech/tags/events.md>), [germany](<https://devfeed.tech/tags/germany.md>), [industry-conference](<https://devfeed.tech/tags/industry-conference.md>), [meetup](<https://devfeed.tech/tags/meetup.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [programming](<https://devfeed.tech/tags/programming.md>), [software](<https://devfeed.tech/tags/software.md>), [software-development](<https://devfeed.tech/tags/software-development.md>), [students](<https://devfeed.tech/tags/students.md>), [systems](<https://devfeed.tech/tags/systems.md>), [systems-programming](<https://devfeed.tech/tags/systems-programming.md>), [technology](<https://devfeed.tech/tags/technology.md>), [zephyr](<https://devfeed.tech/tags/zephyr.md>)

### AI overview

The Zephyr Project meetup is scheduled for November 13, 2026, from 9 a.m. to noon at electronica 2026 in Munich. Its agenda covers practical Zephyr work: OTA updates and their integration and reliability challenges; a microcontroller snow predictor using TensorFlow Lite Micro; tradeoffs in Kconfig and Devicetree, including compile-time hardware configuration; and security and Cyber Resilience Act readiness through upstream maintenance, MCUboot, secure updates, and boot-chain protections. Registration is required through edge lab LIVE!, which manages the event; the description says admission and a one-day electronica pass are free.

### Source excerpt

The Zephyr Project community is heading to Munich, Germany! Join us on Friday, November 13, 2026, from 9:00 AM-12:00 PM for an in-person Zephyr Project Meetup, co-located with edge lab LIVE! at electronica 2026. The event is open to developers, engineers, contributors, students, technology professionals, and anyone interested in open source, embedded systems programming, and software development.

## Implementing Keyword Spotting on the MCXN947 Microcontroller with Integrated NPU

DevFeed: [Implementing Keyword Spotting on the MCXN947 Microcontroller with Integrated NPU](<https://devfeed.tech/articles/implementing-keyword-spotting-on-the-mcxn947-microcontroller-with-integrated-npu-120356.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2506.08911>)

Author: Petar Jaku\\v{s}, Hrvoje D\\v{z}apo

Published: 2026-11-06T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Embedded Systems](<https://devfeed.tech/topics/embedded-systems.md>), [Microcontroller](<https://devfeed.tech/topics/microcontroller.md>), [tinyml](<https://devfeed.tech/topics/tinyml.md>), [quantization](<https://devfeed.tech/topics/quantization.md>)

Tags: [cnn](<https://devfeed.tech/tags/cnn.md>), [cs-ar](<https://devfeed.tech/tags/cs-ar.md>), [cs-hc](<https://devfeed.tech/tags/cs-hc.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cs-pf](<https://devfeed.tech/tags/cs-pf.md>), [cs-sd](<https://devfeed.tech/tags/cs-sd.md>), [embedded](<https://devfeed.tech/tags/embedded.md>), [inference](<https://devfeed.tech/tags/inference.md>), [keyword-spotting](<https://devfeed.tech/tags/keyword-spotting.md>), [leveraging](<https://devfeed.tech/tags/leveraging.md>), [low-power](<https://devfeed.tech/tags/low-power.md>), [microcontroller](<https://devfeed.tech/tags/microcontroller.md>), [npu](<https://devfeed.tech/tags/npu.md>), [quantization](<https://devfeed.tech/tags/quantization.md>), [real-time](<https://devfeed.tech/tags/real-time.md>)

### AI overview

The paper reports a keyword spotting system for the NXP MCXN947 that pairs MFCC feature extraction with a CNN classifier. Quantization Aware Training reduces model size with minimal reported accuracy loss. In experiments, using the integrated NPU made inference 59 times faster than CPU-only execution; the model reached 97.06% accuracy at 30.58 KB. These results suggest the system is feasible for real-time voice interaction on resource-constrained embedded devices.

### Source excerpt

arXiv:2506.08911v2 Announce Type: replace-cross Abstract: This paper presents a keyword spotting (KWS) system implemented on the NXP MCXN947 microcontroller with an integrated Neural Processing Unit (NPU), enabling real-time voice interaction on resource-constrained devices. The system combines MFCC feature extraction with a CNN classifier, optimized using Quantization Aware Training to reduce model size with minimal accuracy drop. Experimental results demonstrate a 59x speedup in inference time when leveraging the NPU compared to CPU-only execution, achieving 97.06% accuracy with a model size of 30.58 KB, demonstrating the feasibility of efficient, low-power voice interfaces on embedded platforms.

## Energy Measurement and Optimization of TinyML in Resource-Constrained Embedded Systems: A Systematic Mapping Study

DevFeed: [Energy Measurement and Optimization of TinyML in Resource-Constrained Embedded Systems: A Systematic Mapping Study](<https://devfeed.tech/articles/energy-measurement-and-optimization-of-tinyml-in-resource-constrained-embedded-systems-a-systematic-mapping-study-120591.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.07264>)

Author: Clariele Almeida, Renan Beserra, Rafael Moura, Danilo Ara\\'ujo, Bruno Nogueira, S\\'ergio Chevtchenko, Ermeson Andrade

Published: 2026-11-06T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [tinyml](<https://devfeed.tech/topics/tinyml.md>), [Embedded Systems](<https://devfeed.tech/topics/embedded-systems.md>), [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [Sustainable Computing](<https://devfeed.tech/topics/sustainable-computing.md>), [reproducibility](<https://devfeed.tech/topics/reproducibility.md>)

Tags: [cs-ar](<https://devfeed.tech/tags/cs-ar.md>), [cs-et](<https://devfeed.tech/tags/cs-et.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [embedded](<https://devfeed.tech/tags/embedded.md>), [energy](<https://devfeed.tech/tags/energy.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [mapping](<https://devfeed.tech/tags/mapping.md>), [measurement](<https://devfeed.tech/tags/measurement.md>), [optimization](<https://devfeed.tech/tags/optimization.md>), [tinyml](<https://devfeed.tech/tags/tinyml.md>)

### AI overview

The mapping finds that energy-aware TinyML research commonly evaluates convolutional and lightweight models using direct measurements or analytical estimates. Energy use is associated with model architecture, hardware, communication, numerical precision, and input characteristics. Reported optimization approaches fall into four groups: model compression, hardware-aware design, system and communication optimization, and data and processing optimization. The abstract also identifies limited standardization in platforms, measurement procedures, and reporting, which constrains reproducibility and direct comparison across studies.

### Source excerpt

arXiv:2611.07264v1 Announce Type: new Abstract: The rapid expansion of the Internet of Things (IoT) has increased the volume of data generated at the network edge, where many devices operate under severe memory, processing, and energy constraints. In this context, Tiny Machine Learning (TinyML) enables machine learning (ML) tasks to be executed on microcontrollers, smart sensors, and other resource-constrained platforms. However, the literature remains heterogeneous regarding the algorithms evaluated, energy-measurement methods, reported metrics, influencing factors, and optimization strategies. This study presents a systematic mapping of energy-aware TinyML research. It identifies the main models and learning approaches, classifies energy-evaluation methodologies and metrics, examines the technical factors affecting energy consumption, and organizes the reported optimization techniques into a structured taxonomy. The findings highlight the frequent use of convolutional and lightweight architectures, direct measurements and analytical estimations, and factors related to model architecture, hardware, communication, numerical precision, and input characteristics. The optimization strategies were grouped into model compression, hardware-aware design, system and communication optimization, and data and processing optimization. The results also reveal limited standardization in experimental platforms, measurement procedures, and reporting practices, limiting reproducibility and direct comparison across studies. This mapping provides an updated overview of the field and guidance for developing and evaluating more energy-efficient TinyML solutions.

## Enabling AWS Cloud FPGA Development Using High-Level Vitis HLS

DevFeed: [Enabling AWS Cloud FPGA Development Using High-Level Vitis HLS](<https://devfeed.tech/articles/enabling-aws-cloud-fpga-development-using-high-level-vitis-hls-118512.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.06829>)

Author: David Rostcheck

Published: 2026-11-05T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [cloud-infrastructure](<https://devfeed.tech/topics/cloud-infrastructure.md>), [Amazon EC2](<https://devfeed.tech/topics/amazon-ec2.md>), [tinyml](<https://devfeed.tech/topics/tinyml.md>)

Tags: [aws](<https://devfeed.tech/tags/aws.md>), [aws-ec2](<https://devfeed.tech/tags/aws-ec2.md>), [circuit](<https://devfeed.tech/tags/circuit.md>), [closure](<https://devfeed.tech/tags/closure.md>), [cloud](<https://devfeed.tech/tags/cloud.md>), [coding-agents](<https://devfeed.tech/tags/coding-agents.md>), [cpu](<https://devfeed.tech/tags/cpu.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-ar](<https://devfeed.tech/tags/cs-ar.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [debugging](<https://devfeed.tech/tags/debugging.md>), [deployment](<https://devfeed.tech/tags/deployment.md>), [development](<https://devfeed.tech/tags/development.md>), [embedding](<https://devfeed.tech/tags/embedding.md>), [engineering](<https://devfeed.tech/tags/engineering.md>), [errors](<https://devfeed.tech/tags/errors.md>), [fpga](<https://devfeed.tech/tags/fpga.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [hls](<https://devfeed.tech/tags/hls.md>), [implementation](<https://devfeed.tech/tags/implementation.md>), [layer](<https://devfeed.tech/tags/layer.md>), [vitis](<https://devfeed.tech/tags/vitis.md>)

### AI overview

The paper presents an end-to-end workflow for developing ML designs on AWS EC2 F2 cloud FPGAs, bridging Vitis HLS through the HDK to AFI deployment. It proposes four validation levels--FP32 PyTorch, bit-accurate quantized Python, HLS C-simulation, and physical FPGA--to help locate errors, including circuit-specific ones. The authors describe deployment, timing closure, and practical HLS decisions, and demonstrate the workflow on TinyBERT's embedding layer and first encoder layer for address named entity recognition. Their design streams weights from host DRAM and performs poorly compared with CPU/GPU implementations; the abstract says they discuss possible forward paths.

### Source excerpt

arXiv:2611.06829v1 Announce Type: cross Abstract: We present a toolchain for development of ML models on AWS EC2 F2 cloud FPGA infrastructure from high-level language, implementing a complete Vitis HLS to HDK bridge to AFI workflow that has been absent from the academic literature despite eight years of AWS FPGA availability. We introduce a four-level methodology (FP32 PyTorch, bit-accurate quantized Python, HLS C-simulation, physical FPGA) to aid in localizing errors, including FPGA-specific circuit errors. Together these tools allow coding agents to effectively implement FPGA designs from user directions, testing and debugging as needed. We document the deployment workflow, timing closure strategies, and practical HLS engineering decisions that proved crucial. We validate using TinyBERT fine-tuned for address Named Entity Recognition, implementing the embedding layer and the first of four encoder layers on the FPGA hardware. Our implementation streams weights from host DRAM, yielding poor performance compared to CPU/GPU implementations; we discuss forward paths.

## Edge AI and TinyML for Resource-Constrained Agriculture: Hardware Platforms, Model Optimization, and Deployment Strategies

DevFeed: [Edge AI and TinyML for Resource-Constrained Agriculture: Hardware Platforms, Model Optimization, and Deployment Strategies](<https://devfeed.tech/articles/edge-ai-and-tinyml-for-resource-constrained-agriculture-hardware-platforms-model-optimization-and-deployment-strategies-116677.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2603.15085>)

Author: Riya Samanta, Bidyut Saha, Soumya Kanti Ghosh

Published: 2026-11-04T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [tinyml](<https://devfeed.tech/topics/tinyml.md>), [quantization](<https://devfeed.tech/topics/quantization.md>), [On-device AI](<https://devfeed.tech/topics/on-device-ai.md>), [AI Integration Strategies](<https://devfeed.tech/topics/ai-integration-strategies.md>), [Embedded Systems](<https://devfeed.tech/topics/embedded-systems.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>)

Tags: [accelerators](<https://devfeed.tech/tags/accelerators.md>), [agriculture](<https://devfeed.tech/tags/agriculture.md>), [ai](<https://devfeed.tech/tags/ai.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [compute](<https://devfeed.tech/tags/compute.md>), [connectivity](<https://devfeed.tech/tags/connectivity.md>), [cs-et](<https://devfeed.tech/tags/cs-et.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [deployment](<https://devfeed.tech/tags/deployment.md>), [deployment-strategies](<https://devfeed.tech/tags/deployment-strategies.md>), [edge](<https://devfeed.tech/tags/edge.md>), [edge-ai](<https://devfeed.tech/tags/edge-ai.md>), [energy](<https://devfeed.tech/tags/energy.md>), [esp32](<https://devfeed.tech/tags/esp32.md>), [farming](<https://devfeed.tech/tags/farming.md>), [flash](<https://devfeed.tech/tags/flash.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [model](<https://devfeed.tech/tags/model.md>), [optimization](<https://devfeed.tech/tags/optimization.md>), [platforms](<https://devfeed.tech/tags/platforms.md>), [resource](<https://devfeed.tech/tags/resource.md>), [tinyml](<https://devfeed.tech/tags/tinyml.md>)

### AI overview

The study finds that precision-farming Edge AI commonly places inference on local devices, while training and model lifecycle management remain centralized. Microcontrollers such as ESP32, STM32, and ATmega are prominent endpoint platforms; single-board computers and embedded accelerators serve more compute-intensive workloads. In the subset of studies comparing optimization methods, almost 50% report quantization, while structured pruning, hardware-aware architecture search, and cross-layer optimization are less common. Comparisons remain difficult because flash, peak RAM, MAC count, latency, and energy per inference are not reported consistently.

### Source excerpt

arXiv:2603.15085v2 Announce Type: replace-cross Abstract: Deploying artificial intelligence at the edge is increasingly important for systems that must operate under strict limits on memory, computation, energy, connectivity, and hardware cost. Tiny Machine Learning (TinyML) represents the most resource-constrained regime of this shift, enabling inference on microcontroller-class devices with tightly bounded flash and SRAM budgets. This study examines recent deployment trends in Edge AI and TinyML using precision farming as a representative real-world environment in which sensing nodes, gateways, UAVs, and embedded processors must often operate with intermittent connectivity and limited power. The analysis emphasizes the deployment stack rather than predictive accuracy alone, covering hardware classes, software toolchains, model- and system-level optimization, intelligence placement, and hardware-resource reporting. Across the analyzed studies, microcontroller-class platforms such as ESP32, STM32, and ATmega are prominent endpoint targets, while single-board computers and embedded accelerators support more compute-intensive workloads. Among the studies included in the optimization comparison, almost 50% report quantization, making reduced-precision inference the most frequently identified strategy within this subset. In contrast, structured pruning, hardware-aware neural architecture search, and coordinated cross-layer optimization remain comparatively limited. Resource profiling is also inconsistent as flash, peak RAM, MAC count, latency, and energy per inference are not reported uniformly, limiting reproducibility and comparisons across deployments. A recurring systems-level pattern is an inference-training asymmetry in which inference moves toward local devices while training and model lifecycle management remain predominantly centralized.

## HyVox: Rethinking Low-Power Voice Intelligence towards Neuromorphic-von Neumann Edge Computing

DevFeed: [HyVox: Rethinking Low-Power Voice Intelligence towards Neuromorphic-von Neumann Edge Computing](<https://devfeed.tech/articles/hyvox-rethinking-low-power-voice-intelligence-towards-neuromorphic-von-neumann-edge-computing-116838.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.03735>)

Author: Snehasis Addy, Eduardo Calle Ortiz, Prashant Shenoy, Phuc Nguyen

Published: 2026-11-04T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>), [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Neural Network](<https://devfeed.tech/topics/neural-network.md>), [tinyml](<https://devfeed.tech/topics/tinyml.md>), [On-device AI](<https://devfeed.tech/topics/on-device-ai.md>), [cosine similarity search](<https://devfeed.tech/topics/cosine-similarity-search.md>), [quantization](<https://devfeed.tech/topics/quantization.md>)

Tags: [2](<https://devfeed.tech/tags/2.md>), [3](<https://devfeed.tech/tags/3.md>), [abstract](<https://devfeed.tech/tags/abstract.md>), [abstraction](<https://devfeed.tech/tags/abstraction.md>), [ann](<https://devfeed.tech/tags/ann.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [cast](<https://devfeed.tech/tags/cast.md>), [chip](<https://devfeed.tech/tags/chip.md>), [compress](<https://devfeed.tech/tags/compress.md>), [computing](<https://devfeed.tech/tags/computing.md>), [cosine](<https://devfeed.tech/tags/cosine.md>), [cosine-similarity](<https://devfeed.tech/tags/cosine-similarity.md>), [counts](<https://devfeed.tech/tags/counts.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cs-ne](<https://devfeed.tech/tags/cs-ne.md>), [cs-sd](<https://devfeed.tech/tags/cs-sd.md>), [edge](<https://devfeed.tech/tags/edge.md>), [edge-computing](<https://devfeed.tech/tags/edge-computing.md>), [intelligence](<https://devfeed.tech/tags/intelligence.md>), [low-power](<https://devfeed.tech/tags/low-power.md>), [neuromorphic](<https://devfeed.tech/tags/neuromorphic.md>), [power](<https://devfeed.tech/tags/power.md>), [voice](<https://devfeed.tech/tags/voice.md>)

### AI overview

HyVox divides open-set speaker identification between a neuromorphic processor that creates speaker embeddings and a microcontroller that stores the speaker gallery, scores embeddings by cosine similarity, rejects unknown speakers, and enrolls users by adding prototypes without updating network weights. Its 832-neuron temporal SNN backbone feeds a 32-neuron spiking embedding extractor. To meet chip constraints, the backbone is compressed 6.5x to 29,000 synapses and trained for integer dynamics; chip-accurate spiking training uses a differentiable replica of the chip's integer updates and doubles mapped-backbone top-1 accuracy on unseen speakers. Across 20 Common Voice households, reported top-1 accuracy is 99.6% with two household members and 94.9% with 20; rejection EER is 5.5% and 13.2%, respectively. A 3-second input takes 0.96 seconds and 1.63 mJ dynamic energy. The authors report 42% lower latency than the fastest embedded ANN baseline and 13.5x lower dynamic and 5.8x lower total energy than the most energy-efficient baseline.

### Source excerpt

arXiv:2611.03735v1 Announce Type: cross Abstract: We present HyVox, a hybrid neuromorphic--von Neumann system for low-energy open-set speaker identification under tight energy budgets. HyVox splits the workload at the speaker embedding. A neuromorphic processor embeds the speech with sparse, event-driven computation, and a von Neumann microcontroller holds the speaker gallery and makes every decision: it scores queries by cosine similarity, rejects unknown speakers, and enrolls new users by writing a prototype, with no weight update. On the neuromorphic processor, a hierarchical temporal abstraction (HTA) SNN backbone learns a synaptic time constant per leaky integrate-and-fire (LIF) neuron, allocating 832 neurons across fast, medium, and slow timescales. A 32-neuron spiking embedding extractor (SEE), pretrained on episodes that each enroll new speakers and reject outsiders, turns the backbone output into an embedding for speakers unseen in training; its 32 spike counts are the only interface between the two processors. To fit the chip's synapse and integer budgets, we compress the backbone 6.5x to 29.0k synapses, train it for integer bit-shift dynamics, and fine-tune it with chip-accurate spiking training (CAST). CAST trains through a differentiable replica of the chip's integer neuron update, derived from its cycle-accurate simulator, so that gradient descent optimizes the exact weights and thresholds the chip runs; it doubles the top-1 accuracy of the mapped backbone on unseen speakers. On 20 Common Voice households of unseen speakers, HyVox reaches 99.6\% top-1 and 5.5\% rejection EER with 2 members and 94.9\% and 13.2\% with 20, and the network deployed on the neuromorphic chip matches its trained model. Each inference on a 3-s window takes 0.96 s and 1.63 mJ of dynamic energy, 42\% lower latency than the fastest embedded ANN baseline and 13.5x less dynamic and 5.8x less total energy than the most energy-efficient one.

[Next page](<https://devfeed.tech/topics/tinyml.md?cursor=WyIyMDI2LTExLTA0VDA1OjAwOjAwKzAwOjAwIiwgIjliYzFjNWI5LTE4ZTItNDc3NS1iMDQ2LTU5Mjc0MGY1YmEzMyJd>)