Title: Enhanced control of a brain–computer interface by tetraplegic participants via neural-network-mediated feature extraction

• Ref: https://www.nature.com/articles/s41551-024-01297-1
• Summary: To construct a high-quality signal, BCI must extract features that accurately estimate neural activity (i.e., neuronal action potential signal). Feature-engineering techniques to recover functional information have advanced with the advancement of AI, and the quantified goal is a high SNR. A CNN combining feature extraction and decoding can be used to map neural activity. This neural network led to excellent performance both offline and online in a cursor-control task.