Automated System Design for the Identification of Sleep Disorder: Cross-correlation and SVM Based Approach

Abstract

In this research work, an automated system is developed for the identification of sleep disorder by analyzing Electroencephalograph (EEG) signals. The EEG signals considered in this study are taken from the Physionet database. The signals have been recorded during the sleeping time of various healthy and unhealthy patients, having sleep disorder. The paper introduces a protocol of feature extraction, involving cross-correlation. The cross-correlation operation automatically eliminates the noises contaminating the electroencephalograph (EEG) signals. The features extracted from the cross-correlograms, besides containing some traditional and not so common parameters, also includes the Higuchi’s Fractal Dimension (HFD). The extracted features from Cross-correlation are processed using Support Vector Machine (SVM), which gives an acceptable accuracy as compare to other research works in bio-signal processing field. The Proposed methodology has achieved 96.65% sensitivity, 100% specificity and 96.67% accuracy. Thus the proposed scheme may be a strong candidate for embedded system applications, where it can be implemented using microcontrollers.