VO₂ Memristor-Based Adaptive Neurons for Electromyography Signal Processing

Hardware implementations of bio-plausible neurons are highly valuable for developing sparse and efficient neuromorphic computing systems. However, traditional CMOS-based neurons require complex auxiliary circuits and bulky capacitors. Here, we experimentally demonstrate a VO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> memristor-based adaptive leaky integrate-and-fire (ALIF) neuron that enables advanced spike-frequency adaptation with ultra-low hardware cost (2T2R1C). These ALIF neurons serve as neural encoders, efficiently extracting temporal signal features with a sparse-spiking yet high-fidelity method. The spiking neural network (SNN) with ALIF achieves 97.1% accuracy in electromyography classification, showing great potential in human-computer interaction.