This is a repository for source code associated with a signal processing research project at Michigan State University's Advanced Microsystems and Circuits (AMSaC) lab, homepage at http://www.egr.msu.edu/amsac. It consists of a software prototype/proof of concept (Matlab), digital hardware (Verilog HDL) and testing/supplementary scripts (various languages) for a real-time neural action potential classification algorithm, as described in the paper

Y. Yang, C. S. Boling, A. J. Mason, "Power-area efficient VLSI implementation of decision tree based spike classification for neural recording implants", IEEE BioCAS 2014.

Extracellular recordings of neural data using intracortical microelectrodes have shown promise for medical and rehabilitative neuroscience applications, e.g. brain-machine interfaces. The principal complications with this approach are 1) massive bandwidth and power requirements as systems scale to the hundreds/thousands of channels required for next-generation neurotechnologies and 2) determining which individual cell is responsible for producing a specific recorded action potential ("spike") in the neural time series. This work attempts to improve the power performance of a parallel digital hardware approach to the latter problem, simultaneously reducing the system bandwidth requirements by transmitting only neuron IDs.

For more details and related work please visit http://www.egr.msu.edu/amsac/nsp.htm.