[16a-M_123-1~13] FS.1 Focused Session "AI Electronics"

12:15 PM - 12:30 PM JST(3:15 AM - 3:30 AM UTC)

[16a-M_123-13] Speech and Character Recognition via Reservoir Computing Using Type-II Superconductivity

〇(D)Ken Arita¹, Jukiya Kusuki¹, Edmund Soji Otabe¹, Yuki Usami¹, Ahmet Karakari¹, Muzhen Xu¹, Hirofumi Tanaka¹, Tetsuya Matsuno² (1.Kyushu Inst. Tech., 2.NIT-Aritake)

Keywords:

Superconductor,Reservoir Computing,Digit Classification

This study evaluated the effectiveness of reservoir computing using type-II superconductors for speech and character recognition. Numerical simulations were performed by solving the Time-Dependent Ginzburg-Landau (TDGL) equations , using pre-processed speech and MNIST time-series data as inputs. By using local electric fields as reservoir outputs , accuracies of 76.4% for speech recognition and 70.0% for character recognition were achieved, demonstrating high utility in digit classification tasks.