[17p-S4_202-1~12] 9.5 New functional materials and new phenomena

4:00 PM - 4:15 PM JST(7:00 AM - 7:15 AM UTC)English Presentation

[17p-S4_202-8] Japanese Syllabary Recognition via YMnO₃ Single Crystal-based Physical Reservoir Computing

〇Muzhen Xu¹, Aoi Terada², Koki Nietani², Yoichi Horibe^1,3, Hirofumi Tanaka^1,2 (1.Kyushu Inst. Tech. Neumorph Center, 2.Kyushu Inst. Tech. Hum. Intel. Sys., 3.Kyushu Inst. Tech. Mater. Sci. Eng.)

Keywords:

physical reservoir computing,multiferroic single crystal,voice recognition

Physical reservoir computing (PRC) has emerged as a hardware-efficient alternative to software-based neural networks, utilizing the nonlinear dynamics of physical substrates to process temporal data. Multiferroic materials with intriguing domain/wall structures are particularly suitable for PRC due to their inherent nonlinear behavior and relaxation properties.[1-2] In this study, we investigate the performance of YMnO3 single crystals as a physical reservoir for speech recognition. Specifically, we evaluate the system's ability to classify 46 Japanese syllables by comparing a direct 46-class independent classification approach against a phonetic structured approach (separating vowel and consonant components).