Open AccessEEG signal detection and analysis with application in educational robotics
Author(s) -
Adam Henrique Moreira Pinto,
Roseli Aparecida Francelin Romero
Publication year - 2020
Language(s) - English
Resource type - Conference proceedings
DOI - 10.5753/wtdr_ctdr.2020.14951
Subject(s) - computer science , artificial intelligence , convolutional neural network , feature extraction , signal (programming language) , wavelet , wavelet transform , pattern recognition (psychology) , electroencephalography , robot , humanoid robot , perceptron , robotics , signal processing , multilayer perceptron , fast fourier transform , short time fourier transform , fourier transform , artificial neural network , digital signal processing , fourier analysis , algorithm , mathematics , psychology , psychiatry , programming language , computer hardware , mathematical analysis
Brain-Computer Interfaces add information to robots directly from users’ brain, allowing for the interpretation of attention, engagement, and even student mistakes. However, most applications still have low accuracy in recognizing this information. In this paper, an Error Related Potential (ErrP) detection system is being proposed. For this, a new database was created by using a serious game and a humanoid robot aiming to force errors and mental state changings of the user. Wavelets and Fourier Transforms were compared to signal feature extraction, classified using both MultiLayer Perceptron (MLP) and Convolutional Neural Networks (CNN). Experiments demonstrate that the wavelet outperformed Fourier transform to extract the ErrP signal, and CNN had a higher accuracy than MLP in the classification.