Classification of Individual Finger Movements from ECoG Signals using a Spiking Neural Network

We present the first classifier based on a spiking neural network (SNN) that can decode individual finger movements from electrocorticography (ECoG) signals. The SNN has only six leaky integrate-and-fire neurons and uses carefully selected features: the local motor potential and the high gamma band power to analyse a publicly available ECoG dataset. Through the investigation of the key aspects affecting SNN performance (epoch length, lag time, number of concatenated epochs), the presented decoder achieves a 72.4% average classification accuracy across three subjects with an average training time of 3.55 s and a latency of only 1 ms. This work demonstrates how a simple SNN architecture can effectively decode complex motor intentions from ECoG signals, potentially enabling more efficient brain-computer interfaces.