MEMS-Based High-Resolution Neural Interfaces for Chinese Phonemes Decoding

We present a high-resolution, flexible neural interface based on microelectromechanical systems (MEMS) for electrocorticographic (ECoG) recording and speech decoding. Owing to the high spatial density and excellent conformal coverage of the electrode arrays, discriminative neural patterns of Chinese vowels were successfully recorded and extracted, revealing spatial encoding characteristics in cortical activity. With customized neural network-based decoders, the proposed speech braincomputer interface (BCI) system achieved a high decoding accuracy with only 6 minutes of intraoperative data. Our method demonstrates the potential for Chinese speech decoding with the ECoG signals, offering the possibility of MEMS-base high-resolution neural interfaces for speech neuroprostheses.