This chapter presents a novel hardware architecture for correlation-based spike detection and unsupervised clustering. The architecture is able to utilize the information extracted from the results of spike clustering for efficient spike detection. The architecture supports the fast computation for the normalized correlation and OSORT operations. The normalized correlation is used for template matching for accurate spike detection. The OSORT algorithm is adopted for unsupervised classification of the detected spikes. The mean of spikes of each cluster produced by the OSORT algorithm is used as the templates for subsequent detection. The architecture adopts postnormalization technique for reducing the area costs. Modified OSORT operations are also proposed for facilitating unsupervised clustering by hardware. The proposed architecture is implemented by field programmable gate array (FPGA) for performance evaluation. In addition to attaining high detection and classification accuracy for spike sorting, experimental results reveal that the proposed architecture is an efficient design providing low area cost and high throughput for real-time offline spike sorting applications.
Part of the book: Field