Energy Efficient Spiking Temporal Encoder Design for Neuromorphic Computing Systems

Neuromorphic computing hardware has undergone a rapid development and progress in the past few years. One of the key components in neuromorphic computing systems is the neural encoder which transforms sensory information into spike trains. In this paper, both rate encoding and temporal encoding schemes are discussed. Two novel temporal encoding schemes, parallel and iteration, are presented. The power consumption of the encoder has been significantly reduced by combing the iteration encoding and low sampling rate in advanced complementary metal-oxide semiconductor (CMOS) nano-technology. Both the simulation and measurement results show the accuracy and efficiency of the proposed encoding circuits. The proposed iteration encoder has immediate applicability as a general purpose input encoder for a reservoir computing system.