A configurable qualitative-modeling-based silicon neuron circuit

Abstract

A silicon neuronal network is a neuro-mimetic system that aims to realize an electronic-circuit version of the nervous system by connecting silicon neuron circuits via silicon synapse circuits. In our previous works, we proposed a qualitative-modeling-based design approach that provides a solution to the trade off in the silicon neuron circuits between the power consumption and the variety of supported neuronal activities. By this approach, we developed an analog silicon neuron circuit that can be configured to Class I, Class II, regular spiking, elliptic bursting, and square-wave bursting modes with power consumption less than 72 nW. Simulation and experimental results for the first 4 modes are reported.