On-Chip Spike Clustering & Classification using Self Organizing Map for Neural Recording Implants

Spike sorting plays a vital role in recording neural signals using microelectrode arrays for neuroscience research. The bandwidth bottleneck is greatly eased for data transmission of hundreds of channels in neuroprosthetic application through on-chip spike sorting. Our previous work on feature extraction algorithm called Zero-Crossing Features (ZCF) shows good performance for data reduction and spike classification while requiring minimal hardware resources. In this paper, a new spike clustering & classification method based on ZCF feature extraction is presented. The method is shown to perform well in identifying the number of neurons in a spike channel and classifying spikes into correct neurons. An implant-compatible VLSI hardware architecture to cluster and classify spikes in the ZCF feature space is also presented.