Time-to-First-Spike Coding in Sensory Neuron for Energy-Efficient Spiking Neural Networks

A neuromorphic visual system based on artificial sensory neurons offers a promising way to reduce power consumption in image sensing and processing. In this study, an inhibitive sensory neuron capable of time-to-first-spike (TTFS) coding is proposed, designed to function as input neurons in a spiking neural network. A floating body MOSFET is used as the sensory neuron, where the gate voltage can dynamically switch an excited state to an inhibited state after generating the first spike. In handwritten digit recognition, TTFS coding reduces the total spike count per inference by 89.9% compared to rate coding, significantly enhancing the energy efficiency of the visual system.