Performance Evaluation of FPGA based Processor for Pattern Recognition of EEG

Abstract

Brain-machine interface (BMI) is a technology that enables operation for electromechanical devices using biological signals such as EEG or MEG. BMI is enables real-time communication with others for patients such as ALS. In the BMI field, pattern recognition of EEG is required to be processed in high precision and online. As a method, machine learning or deep learning have been used. However, when high accuracy is required, the amount of calculation becomes enormous, and online processing becomes difficult.Therefore, we have been developed the dedicated processor for pattern recognition of EEG in real-time. If it can be implemented in one chip, it is useful for realizing the advanced BMI devices. Hence, we have been designed the inference processor based on EEGNet. We report the results of performance evaluation such as accuracy of pattern recognition and execution time when the designed processor implemented in one chip FPGA.