FPGA based real-time detection of spikes within neural signals

This paper presents a technique for real time multi-channel signal processing of biological neural data. The major objective is the extraction of action potentials from neural data streams continuously acquired from a multi-electrode sensor array. For the actual detection process a wavelet-based method will be proposed, which allows efficient filtering as well as the definition of a robust threshold. In order to match the real time constraints for the wavelet transform an efficient and fast algorithm (lifting scheme) was used. An exemplary 8-channel implementation will be described exploiting the parallel potentials of a Xilinx Virtex field programmable gate array (FPGA). This work is part of a project concerned with the development of a microsensorchip system for neural data analysis at the University of Rostock, Germany.