Design of a Dedicated Hardware Architecture for Biosignal Processing

Abstract

The extraction of electrical information from bio-siognal recordings is been a field where the Action Potential Detection and Sorting process take place. The analysis when this information is detected is crucial for \tbi{in vivo} and \tbi{in vitro} recordings, as well as the time that it takes to detect such electrical information and their analysis in off-line process. In addition, this off-line analysis could take along time and many memory resources. Therefore, we proposed a simple method to achieve Detection and Classification of Action Potentials by using a match filter technique combined with the detection of Correlation Pattern by an adaptive threshold in a hardware architecture. The architecture was realized in a Field-Programable Gate Array (FPGA) by VHDL language, in order to achieve the goal of real-time performing and be a good candidate in both real-time and off-line process. The architecture consist of a chain module that can be replicated for Micro electrode arrays (MEAs) recordings with some adjusments. The architecture is able to save the sample number of the Action potential detected and sorting this sample in six different cluster depending on the Correlation Pattern detected. This Correlation Pattern is the result of the correlation with the mean Action Potential template and Action Potential template detected at that moment. The VHDL module have an Universal asynchronous receiver-transmitter UART communication protocol to start the process, stop it and read only a sample at the time. An additional feature is added in the off-line mode, where each time an Action Potential is detected, the module stop the reading process in the SD-card and the architecture, and the Action Potential shape, thresholds and Correlation shape can be seen on the OLED display. The chain module was tested with a Macaque monkey biosignal recorded in vivo at a sampling rate of 40 kHZ and...