Spectral entropy feature subset selection using SEPCOR to detectalcoholic impact on gamma sub band visual event related potentialsof multichannel electroencephalograms (EEG)

Padmashree, T K and Sriraam, N (2017) Spectral entropy feature subset selection using SEPCOR to detectalcoholic impact on gamma sub band visual event related potentialsof multichannel electroencephalograms (EEG). Applied Soft Computing, 46 (46). pp. 441-451. ISSN 1568-4946

[img] PDF
1859.pdf - Published Version
Restricted to Registered users only

Download (2MB) | Request a copy

Abstract

tThe problem of analyzing and identifying regions of high discrimination between alcoholics and controlsin a multichannel electroencephalogram (EEG) signal is modeled as a feature subset selection techniquethat can improve the recognition rate between both groups. Several studies have reported efficient detec-tion of alcoholics by feature extraction and selection in gamma band visual event related potentials (ERP)of a multichannel EEG signal. However, in these studies the correlation between features and their classinformation is not considered for feature selection. This may lead to redundancy in the feature set andresult in over fitting. Therefore in this study, a statistical feature selection technique based on Separa-bility & Correlation analysis (SEPCOR) is proposed to select an optimal feature subset automatically thatpossesses minimum correlation between selected channels and maximum class separation. The optimalfeature selection consists of a ranking method that assigns ranks to channels based on a variability mea-sure (V-measure). From the ranked feature set of highly discriminative features, different subsets areautomatically selected by heuristically applying a correlation threshold in steps from 0.02 to 0.1. Thesesubsets are applied as input features to multilayer perceptron (MLP) neural network and k-nearest neigh-bor (k-NN) classifiers to discriminate alcoholic and control visual ERP. Prior to feature selection, spectralentropy features are computed in gamma sub band (30–55 Hz) interval of a 61-channel multi-trial EEGsignal with multiple object recognition tasks. Independent Component Analysis (ICA) is performed on rawEEG data to remove eye blink, motion and muscle artifacts. Results indicate that both classifiers exhibitexcellent classification accuracy of 99.6%, for a feature subset of 22 optimal channels with correlationthreshold of 0.1. In terms of computation time, k-NN classifier outperforms multilayer perceptron-backpropagation (MLP-BP) network with 7.93 s whereas MLP network takes 55 s to perform the recogni-tion task with the same accuracy. Compared to feature section methods used in previous studies on thesame EEG alcoholic database, there is a significant improvement in classification accuracy based on theproposed SEPCOR method.

Item Type: Article
Uncontrolled Keywords: Visual event related potentials (ERP)Electroencephalogram (EEG)Spectral entropyGamma sub bandSeparab
Subjects: Engineering > MIT Manipal > Electrical and Electronics
Depositing User: MIT Library
Date Deposited: 11 Jan 2017 11:32
Last Modified: 11 Jan 2017 11:32
URI: http://eprints.manipal.edu/id/eprint/148002

Actions (login required)

View Item View Item