Spiking Neural Networks-Inspired Signal Detection Based on Measured Body Channel Response

Spiking neural networks (SNNs) are inspired by biological behavior in the neural system processing information by the rate or delay components of discrete spiking signals in a massively parallel manner. Sparse and asynchronous spikes allow event-driven information processes, leading to low power consumption and fast inference. By exploiting these advantageous features of the SNNs, this article presents a signal detection method for human body communication (HBC), which has recently emerged as an innovative alternative for wireless body area networks using the human body as a signal transmission medium. In particular, binary spike signaling in the SNNs is highly appropriate for application in the digital signal transmission-based HBC systems. The experiments of body channel response (BCR) measurements using digital training signals show that the body channel characteristics vary with changes in body posture and device location, especially in wearable environments requiring small-sized devices powered by batteries. The proposed SNN structures can enhance communication performance from signal distortions, stemming from the effects of the time-dispersive body channel and bandwidth-limited receive filter. The proposed SNN-based transmission symbol code (TSC) detector (STD) can improve about 3.53 dB carrier-to-noise ratio (CNR) at a bit error rate (BER) of 10(-6) for a data rate of 1.3125 Mbps, compared to that of a conventional maximum likelihood (ML) detector. In addition, the proposed SNN-based preamble detector (SPD) can secure an approximately 150 wider threshold range than that of a conventional correlator to achieve a detection probability higher than 99.9% of the frame existence at a CNR of approximately 0 dB required for achieving a BER of 10(-6) by the STD.