Brain-machine interfaces (BMIs) are crucial for restoring lost motor functions and understanding brain mechanisms. However, current BMIs based on von Neumann architecture face limitations in signal processing due to the exponentially increasing number of recording electrodes. This work introduces a memristor-based neural signal analysis system that leverages the bio-plausible characteristics of memristors for efficient analog domain signal processing. Using memristor arrays, the system achieves a 93.46% accuracy in filtering and identifying epilepsy-related neural signals, demonstrating a 400x power efficiency improvement compared to state-of-the-art CMOS systems. This highlights the potential of memristors for high-performance neural signal analysis in next-generation BMIs.

Zhengwu Liu, Jianshi Tang, Bin Gao, Peng Yao, Xinyi Li, Dingkun Liu, Ying Zhou, He Qian, Bo Hong, Huaqiang Wu