Frequency Matching in Spiking Neural Networks for mmWave Sensing
2026-05-11 • Neural and Evolutionary Computing
Neural and Evolutionary Computing
AI summaryⓘ
The authors explore how spiking neural networks (SNNs), which mimic brain activity, can improve millimeter-wave (mmWave) sensing used in edge devices. They found that the natural filtering ability of these networks helps reduce noisy, high-frequency signals, making them better than traditional neural networks (ANNs) when important data lies in lower frequency ranges. By tuning the network's settings to match the signal's frequency, the authors achieved better accuracy and lower energy use on several datasets. Their work shows a way to make mmWave sensing more efficient and accurate without heavy processing.
Millimeter-wave sensingSpiking neural networks (SNNs)Artificial neural networks (ANNs)Leaky integrate-and-fire (LIF) dynamicsFrequency filteringEdge devicesTemporal filteringHigh-frequency noiseEnergy consumptionSignal bandwidth
Authors
Di Yu, Zhenyu Liao, Changze Lv, Wentao Tong, Linshan Jiang, Sijie Ji, Xin Du, Hailiang Zhao, Xiaoqing Zheng, Shuiguang Deng
Abstract
Millimeter-wave (mmWave) sensing enables privacy-preserving, always-on edge perception, but its measurements are often sparse, temporally irregular, and corrupted by high-frequency noise. Existing mmWave pipelines predominantly rely on artificial neural networks (ANNs), which achieve robustness through extensive preprocessing or deep architectures, thereby limiting their efficiency on edge devices. In this work, we study spiking neural networks (SNNs) for mmWave sensing from a mechanism-data alignment perspective. By leveraging the low-pass filtering behavior of leaky integrate-and-fire (LIF) dynamics, we analyze how their implicit temporal filtering interacts with the frequency structure of mmWave signals. Our analysis shows that when discriminative information resides in low-to-mid frequencies, LIF dynamics can inherently suppress high-frequency noise, clarifying when and why SNNs outperform ANNs. Based on this insight, we derive a principled criterion for configuring the membrane decay factor by matching the effective bandwidth of LIF dynamics to the data's discriminative spectral content. Experimental results across four widely used mmWave datasets validate the proposed frequency-matching hypothesis, yielding an average test-accuracy improvement of 6.22% and a 3.64$\times$ reduction in theoretical energy consumption relative to ANN baselines, under a unified evaluation protocol.