Efficient online feature extraction algorithm for spike sorting in a multichannel FPGA-based neural recording system

A novel feature extraction algorithm for multichannel FPGA-based neural recording systems is presented in this paper. It contains the Dual Vertex Threshold (DVT) and the Minimum Delimitation (MD), which are used for spike detection and feature vector extraction respectively. By reducing the computational complexity of DVT and MD, the difficulty of this algorithm in application is greatly reduced. Based on this characteristic, a multichannel FPGA hardware architecture is implemented in this paper. Using extracted feature vectors, the sorting performance of K-means is as good as that with the PCA-based features. Additionally, the test result shows that the transmission bandwidth is reduced to 1.62% of original data rate.