Selection of Acquisition Channels of a Low-Cost Electroencephalogram Device for Signals Based on Motor Imagery for BCI using Typical Testors

Abstract

Brain-computer interfaces based on motor imagination use the intention of movement to communicate with some external device, generally, the signals are acquired, processed, classified, and converted into control commands. However, one of the challenges to establishing interfaces in real-time is the reduction of the dimension of the data, since this increases the delay that exists between the intention of movement and the execution of the action in the device to be controlled. To reduce the delay, this study presents an analysis using typical testors, which are a tool that helps to find patterns in the electroencephalographic signals of motor imagination and thus helps to select the acquisition channels that provide the greatest amount of information. A subject participated in the study, from whom signals of two motor imagination thoughts were acquired, after the analysis of testors, the signals were classified with a recurrent neural network, and the result obtained was 97% accuracy.