[16a-A410-1~8] FS.1 Focused Session "AI Electronics"

[16a-A410-2] Improvement of classification accuracy of speech recognition based on insights into input data characteristics in FeFET reservoir computing with parallel data processing

〇Eishin Nako¹, Kasidit Toprasertpong¹, Ryosho Nakane¹, Mitsuru Takenaka¹, Shinichi Takagi¹ (1.Univ. Tokyo)

Keywords:reservoir computing, FeFET

We have proposed a FeFET reservoir computing with parallel data processing and experimentally demonstrated its performance on a spoken digit recognition task. We achieved a classification accuracy of 95.9% in our previous experiments. We adjust the computing scheme to input data characteristics and achieve the classification accuracy of 98.1%.