An analog VLSI implementation of a multi-scale spike detection algorithm for extracellular neural recordings

This paper discusses a multi-scale neural spike detection algorithm for a low-power analog circuit implementation. The key idea is to implement wavelet decomposition and improve spike detection by independently controlling thresholds for each scale. Each thresholded scale is then combined to provide a single output indicating a spike occurrence. This spike detection algorithm shows promising results towards a robust, compact, and unsupervised low power analog spike detection circuit. A low power frontend spike detection circuit can be added to a neural amplifier and dramatically reduce the required data bandwidth for BMI applications