The Japan Society of Applied Physics

09:30 〜 09:45

[K-7-02] Crystalline Oxide Semiconductor FET-based Analog Neural Network for Intelligent IoT Sensor

〇Hidefumi Rikimaru1, Satoru Ohshita1, Yoshiyuki Kurokawa1, Toshihiko Takeuchi1, Shunpei Yamazaki1 (1. Semiconductor Energy Laboratory Co., Ltd. (Japan))

Presentation style: On-site (in-person)

https://doi.org/10.7567/SSDM.2022.K-7-02

This paper proposes an intelligent Internet of Things (IoT) sensor including an analog neural network (NN) constructed using a crystalline oxide semiconductor field-effect transistor (OSFET). An analog multiply-accumulate (MAC) cell array composed of only OSFETs corresponds to a synapse circuit of each layer in the NN, and monolithically stacked OSFET layers are applicable to a multilayer NN. MAC cells execute MAC operation with current input of the order of picoamperes. Stacking a photoelectric conversion layer on the NN should achieve a low-power intelligent IoT sensor.