Implementation on an FPGA of Perceptron Algorithm for Pattern Classification and Recognition in Electromyographic Digital Signal Processing

Abstract

This paper presents the design and synthesis of a perceptron as an pattern recognition model for its implementation for digital electromyographic signal processing for the classification of movements from wrist and arm. In order to reduce the implementation hardware and accelerate operations for pattern recognition and digital signal processing, the perceptron algorithm, digital signal filtering, segmentation and characteristics extraction are designed with Verilog HDL and implemented at hardware level in a Field Program Gate Array. After obtaining and characterizing the signal, from registered signal values in MATLAB, the simple perceptron training algorithm is used to obtain the synaptic weights of the perceptron, which are then implemented in the FPGA design for the perceptron module in hardware. After implementing the complete system and performing the tests, the signals obtained are analysed with their classification and the percentage of success, with results of 70% and 80% person A and 60% and 80% for persond B for contraction and extension, respectively; it should be noted that the complete system was only tested on two people.