EEG and ECG nonlinear and spectral multiband analysis to explore the effect of videogames against anxiety

Abstract

Currently, the use of video games has purposes that go beyond entertainment and has been gaining prominence in the health area. In this sense, it was hypothesized that it is possible to discriminate biological signals, namely electrocardiographic and electroencephalographic signals, collected from different participants stimulated through three different commercial video games, Tetris, Bejeweled and Energy. To test this hypothesis, a protocol was developed with the Trier Social Stress Test to induce and dose stress in the subjects to similar levels before each game session, in order to observe the effects of the three test games (3 study groups) at the physiological level. Initially collected at 2000 Hz, the signals were resampled to 500 Hz and filtered using a Butterworth low-pass filter. After filtering the signals, several representative features of the study signals were collected. These features consisted of a series of nonlinear metrics such as the Lyapunov exponent and Correlation Dimension, self-similarity metrics such as the Hurst exponent, and detrended fluctuation analysis, fractal dimensions - such as the Katz and Higuchi fractal dimensions - and metrics of signal chaos and activity, such as signal energy, Logarithmic entropy and Shannon entropy, and a number of spectral metrics for the EEG signal, which should be able to help identify any differences in the stress response. As a final result, a discrimination accuracy of 100% was obtained to discriminate the three study groups, using the top 20% of features selected by the F-score technique, using the coarse K Nearest Neighbor classifier.

Date of Award	27 Jul 2022
Original language	English
Awarding Institution	Universidade Católica Portuguesa
Supervisor	Patrícia Oliveira-Silva (Supervisor) & Pedro Miguel Rodrigues (Co-Supervisor)