Analysing EEG brain signals using independent component analysis techniques

Analysing EEG brain signals using independent component analysis techniques

The use of electroencephalography (EEG) in the medical field is evident in the effect it has on diagnosis and treatment of patients who suffer from some form of brain problem. These signals however once collected are overlayed with artifacts. This thesis considers this problem and seeks to solve using popular methods in the form of Independent Component Analysis (ICA) and Wavelet Transform (WT). Independent component analysis (ICA) is a popular blind source separation (BSS) technique that has proven to be promising for the analysis of EEG data. There are different estimators to developing these ICAs. Mutual Information is one of the most natural criteria when
developing an estimator. Although utilized to some level it has always been difficult to calculate. In this thesis I present a new algorithm which utilizes a contrast function related to Mutual Information based on B-Spline functions. This thesis also investigates the creation of an algorithm which is based on a merger of Independent Component Analysis and Translation Invariant Wavelet Transform and goes on to merger the B-Spline ICA with the Translation
Invariant Wavelet Transform. In addition I apply Unscented Kalman Filtering as it does not require any prior signal knowledge. Each algorithm will be examined and compared to ones in literature tackling the same EEG problems; results will be drawn on the base of comparative tests on both synthetic and real.