Wavelets and SVM analysis applied to signals responsive to auditory and visual stimulus captured by EEG in students with ADHD
Abstract— Noninvasive electroencephalographic (EEG) instrumentation has aided in research on human cognitive patterns. In this context, this work proposes an experimental set, including the EEG technique, to investigate patterns of brain signals in learners with Attention Deficit Hyperactivity Disorder (ADHD). The experimental model proposes to evoke the signals during the course of an activity called "Oddball Paradigm". For experimental development, a group of children previously diagnosed with ADHD and a control group were selected. In this article, we will also present the implementation of software for signal acquisition, preprocessing and visualization of the brain signal. The visual and auditory stimulation environment programmed and contemplated in the Oddball Paradigm will be shown. As software support, the mathematical modeling of Wavelet Transform was used for decomposing EEG signals. The data classification method used the Support Vector Machine (SVM) technique to recognize patterns that indicate characteristics of ADHD and normality. The energy and power variables, extracted from the application of the Wavelet Morlet Transform on the EEG records, were used as information for the classifiers. The result of the applied experimental process allowed the construction of a mathematical model able to distinguish individuals with ADHD and control group with 94.74% of accuracy.
Keywords— cognitive pattern disturbances; electroencephalography; wavelets and Morlet; SVM.