Automated Classification of Depression Electro encephalographic Signals using Discrete Cosine Transform and Nonlinear Dynamics

Bairy, Muralidhar G and Bhat, Shreya and Eugene, Lim Wei Jie and Niranjan, U C and Puthankattil, Sudha D and Joseph, Paul K (2015) Automated Classification of Depression Electro encephalographic Signals using Discrete Cosine Transform and Nonlinear Dynamics. Journal of Medical Imaging and Health Informatics, 5 (3). pp. 1-6. ISSN 2156-7026

[img] PDF
05JMIHI03-1418 G M Bairy.pdf - Published Version
Restricted to Registered users only

Download (151kB) | Request a copy

Abstract

Depression is a mental disorder that affects emotional and physical state of a person. It is a state of extreme sadness and dejection. The electroencephalographic (EEG) signals can be used to detect the alterations in the brain’s electrochemical potential. The highly irregular and complex EEG signal variations can be determined by different processing tools. The present work is based on the automated classification of the normal and depression EEG signals. The discrete cosine transform (DCT) decomposes the normal and depression EEG signals into different frequency sub-bands. Nonlinear methods such as sample entropy, correlation dimension, fractal dimension, largest Lyapunov exponent, Hurst exponent and detrended fluctuation analysis are applied to the DCT coefficients and the extracted characteristic features are ranked using t-value. These significant features are fed to decision tree (DT), support vector machine (SVM), k-nearest neighbor (kNN) and naïve Bayes (NB) classifiers. Five significant features are selected and the SVM classifier with radial basis function (RBF) results in a classification accuracy of 93.8%, sensitivity of 92% and specificity of 95.8%.

Item Type: Article
Uncontrolled Keywords: Depression, DCT, Hurst, LLE, DFA, NB.
Subjects: Engineering > MIT Manipal > Biomedical
Depositing User: MIT Library
Date Deposited: 22 Jun 2015 06:25
Last Modified: 22 Jun 2015 06:25
URI: http://eprints.manipal.edu/id/eprint/143107

Actions (login required)

View Item View Item