# Keeping Inference in the Radio: Design and Measurement of a Spiking Accelerator for mmWave ISAC Activity Recognition

DevFeed: [Keeping Inference in the Radio: Design and Measurement of a Spiking Accelerator for mmWave ISAC Activity Recognition](<https://devfeed.tech/articles/keeping-inference-in-the-radio-design-and-measurement-of-a-spiking-accelerator-for-mmwave-isac-activity-recognition-104092.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.37480>)

Author: Anthony S\\'anchez, Rafael Ruiz, Jos\\'e A. Belloch, Jes\\'us O. Lacruz, Joerg Widmer

Published: 2026-10-28T04:00:00Z

Content type: article

Language: en

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

Topics: [Neural Network](<https://devfeed.tech/topics/neural-network.md>), [quantization](<https://devfeed.tech/topics/quantization.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [On-device AI](<https://devfeed.tech/topics/on-device-ai.md>), [model architecture](<https://devfeed.tech/topics/model-architecture.md>)

Tags: [accelerators](<https://devfeed.tech/tags/accelerators.md>), [access-point](<https://devfeed.tech/tags/access-point.md>), [cs-ar](<https://devfeed.tech/tags/cs-ar.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cs-ni](<https://devfeed.tech/tags/cs-ni.md>), [dataset](<https://devfeed.tech/tags/dataset.md>), [design](<https://devfeed.tech/tags/design.md>), [doppler](<https://devfeed.tech/tags/doppler.md>), [energy](<https://devfeed.tech/tags/energy.md>), [framework](<https://devfeed.tech/tags/framework.md>), [fully-connected](<https://devfeed.tech/tags/fully-connected.md>), [generate](<https://devfeed.tech/tags/generate.md>), [handset](<https://devfeed.tech/tags/handset.md>), [ieee](<https://devfeed.tech/tags/ieee.md>), [implement](<https://devfeed.tech/tags/implement.md>), [inference](<https://devfeed.tech/tags/inference.md>), [measurement](<https://devfeed.tech/tags/measurement.md>), [mmwave](<https://devfeed.tech/tags/mmwave.md>), [networks](<https://devfeed.tech/tags/networks.md>), [neural](<https://devfeed.tech/tags/neural.md>), [neural-networks](<https://devfeed.tech/tags/neural-networks.md>), [radio](<https://devfeed.tech/tags/radio.md>), [recognition](<https://devfeed.tech/tags/recognition.md>)

## AI overview

On the public DISC dataset of 60 GHz IEEE 802.11ay measurements, a 132-neuron spiking neural network (SNN) accelerator achieved 94.7% accuracy and 90.7% macro-F1 on unseen subjects, with 206-microsecond inference. Implemented with a spike encoder on a ZCU208 RFSoC, the full path from micro-Doppler frame to activity label runs in programmable logic. Against conventional networks of the same topology on the same device, it used one-eighth the logic and 3.4x less dynamic energy per decision than a network fed the same binary spikes at comparable accuracy; versus one fed a 16-bit spectrogram, it used 7.6x less energy at 1.2 percentage points lower accuracy. Quantization-aware training reduced weights to 3 bits while staying within half a percentage point of the 6-bit network at the same encoder threshold.

## Source excerpt

arXiv:2610.37480v1 Announce Type: cross Abstract: Integrated sensing and communication allows wireless networks to recognize human activity from the channel estimates that the receiver computes anyway. The inference, however, runs on the general-purpose processors of the base station, access point or handset, which are the first components a device switches off to save energy. Always-on sensing therefore requires a classifier that is small enough to remain powered while the rest of the system sleeps. We investigate whether Spiking Neural Networks (SNNs) can provide such a classifier. To this end, we generate SNN accelerators for micro-Doppler activity recognition on the public DISC dataset (IEEE 802.11ay at 60 GHz) with the open-source Spiker+ framework, and implement them together with a spike encoder on a ZCU208 RFSoC, so that the complete chain from micro-Doppler frame to activity label runs in programmable logic. A fully connected SNN with 132 neurons achieves 94.7 % accuracy (90.7 % macro-F1) on unseen subjects and performs an inference in 206 microseconds. Compared to conventional networks of the same topology on the same device, it requires an eighth of the logic and 3.4X less dynamic energy per decision than a network fed the same binary spikes, at comparable accuracy, and 7.6X less than a network fed the 16-bit spectrogram, at 1.2 points lower accuracy. Quantization-aware training reduces its weights to 3 bits within half a percentage point of the 6-bit network at the same encoder threshold