Efficient subspace based techniques for processing single channel electroencephalogram signals

No Thumbnail Available
Journal Title
Journal ISSN
Volume Title
In recent years, the use of portable/wearable EEG systems in health care applications has been increased due its low-power consumption, easy to operate and minimum instrumentation complexity, and reduces the cumbersome to the subject under test. In general, the EEG signals often measured in ambulatory situations, hence, they were contaminated by several artifacts. The presence of these artifacts will degrade the performance of EEG based detection systems. Since the portable EEG devices comprise single or few (at most four) EEG channels, traditional artifact removal techniques, such as blind source separation (BSS), cannot be applied to remove these artifacts. Hence, in this thesis, various subspace based artifact removal techniques for single channel EEG signals were proposed.Singular spectrum analysis (SSA) is a subspace based technique used in this thesis to remove artifacts from single channel EEG signals. Using the frame work of SSA, first, we proposed new grouping criteria to identify the desired signal (EEG signal) subspace. In this criterion, the local mobility of the eigenvectors is considered and employed for removing the motion artifacts from single channel EEG signals. In addition, SSA is combined with ANC to remove the eye blink artifact in EEG epochs recorded for BCI application.
Supervisor: Shaik Rafi Ahamed