Comparing Features Extraction and Classification Methods to Recognize ErrP Signals

Brain Computer Interface systems allow communication between the user and the computer using information provided from brain waves. This control is possible through the interpretation of electrical records and allows the system to analyze the conditions of human attention and engagement in a task. One of the problems in attention during learning is the moment when the student makes a mistake, which is possible to catch through a specific signal of the brain using those systems. However, this data is quite complicated and noisy, and low accuracy was found so far. In this paper, it was compared methods to extract features and classify this signal to enhance the accuracy in error detection. Both wavelets and Fourier transform are used to feature extraction and a MultiLayer Perceptron as compared to Deep Learning Neural Network in classification. The results shown that wavelet extraction are better to extract the error and Deep Learning approach is better to classify.