Wednesday, September 30, 2020 Home   | Search
Viewing Abstract

Page: 52-63 ------------------------------> Last_modified :7/11/2017 12:11:00 PM

Title: Classification Performance of New Fusion Features of P300 in Visual Evoked Potentials from EEG to Distinguish Alcoholics and Controls
Authors: Ravindran Natarajan ; Andrews Samraj
Aff: Department of Computer Applications, Mahendra College of Engineering, Salem, Tamil Nadu, India.
Author Email:
Keywords: EEG; P300; Visual Evoked Potential; KNN; SVM; FFT; SVD; Fractal Dimension; EEG; Na´ve Bayes; Mean Absolute Percentage Error
Abstract:Background: The health factor of alcoholics and controls can be measured significantly using Visually Evoked Potentials (VEP) based on their retort sharpness to visual stimuli. Alcoholism can affect the brain and human performance in assorted ways. People who have the habit of drinking have trouble with their balance, judgment and coordination due to the inflicted brain cells and central nervous system. The predisposition towards alcoholism can be pre detected to avoid it from taking adverse effects among youngsters who abuse alcohol. Objective: If the negative effects of alcoholism are explained to them clearly, with evidence, the habit of abuse can be positively eradicated. A simple visual exercise using pictures is used to extract the P300 component evoked in brain. The extracted P300 varies in dimension among healthy and alcoholic abusers. The Na´ve bayes classifier, KNN clustering and SVM classifier were used on three sets of combinational features constructed by the Fourier features and fractal dimension; Fourier feature and singular value decomposition feature; as well as fractal dimension feature and singular value decomposition from the visual evoked potentials of brain to distinguish the controls from alcohol abusers. This experiment is aimed to take preventive, educative and corrective actions for the benefit of youngsters. The P300 component buried VEP represent in the combo features is expected to provide the substantiation for clustering. Result: It is proved by this research work that the combinational feature vector produced using FFT and FD along with a combination of FD and SVD, as well as FFT and SVD were effective in classifications than the similar classifiers used for simple feature in the previous studies. Conclusion: We proposed the use of combinational features for P300 classification with assorted combinations. This is done with the assurance obtained in classification results of P300 features when used in combination brings a better classification always than the individual feature usage. We concluded with a support to these findings, the trend analysis calculated by the Mean Absolute Percentage Error (MAPE) is also ensures and