Time-Robust and Energy-Efficient Decoder for Real-Time Neural Decoding of Primary Motor Cortex Activity

Our study is to be collocated in the contex of the big European Project B-Cratos. We will explore the potential and efficiency of a new class of neuromorphic algorithms for analyzing brain signals. Specifically, we will develop a Spiking Neural Network (SNN)-based decoder for a Brain-Machine Interface (BMI). The purpose of this decoder will be to interpret 2D hand kinematics from brain signals collected through ECoG technology from the Primary Motor Cortex (M1) of a nonhuman primate, Indy. Additionally, we will investigate both unsupervised and supervised adaptation techniques to efficiently address the evolution in M1's firing dynamics through time. This presents itself a significant challenge because it is both time-consuming and power-intensive, making it unsuitable for real-world applications where efficiency and responsiveness are critical. Finally, we have prepared on simulink a simulation to test real-time scenarios and subsequently program an FPGA to test our decoder.