Signal Processing Techniques for Extracting Signals with Periodic Structure: Applications to Biomedical Signals

In this dissertation some advanced methods for extracting sources from single and mul­ tichannel data are developed and utilized in biomedical applications. It is assumed that the sources of interest have periodic structure and therefore, the periodicity is exploited in various forms. The proposed methods can even be used for the cases where the signals have hidden periodicities, i.e., the periodic behaviour is not detectable from their time representation or even Fourier transform of the signal. For the case of single channel recordings a method based on singular spectrum anal­ ysis (SSA) of the signal is proposed. The proposed method is utilized in localizing heart sounds in respiratory signals, which is an essential pre-processing step in most of the heart sound cancellation methods. Artificially mixed and real respiratory signals are used for evaluating the method. It is shown that the performance of the proposed method is superior to those of the other methods in terms of false detection. More­ over, the execution time is significantly lower than that of the method ranked second in performance. For multichannel data, the problem is tackled using two approaches. First, it is as­ sumed that the sources are periodic and the statistical characteristics of periodic sources are exploited in developing a method to effectively choose the appropriate delays in which the diagonalization takes place. In the second approach it is assumed that the sources of interest are cyclostationary. Necessary and sufficient conditions for extractability of the sources are mathematically proved and the extraction algorithms are proposed. Ballistocardiogram (BCG) artifact is considered as the sum of a number of inde­ pendent cyclostationary components having the same cycle frequency. The proposed method, called cyclostationary source extraction (CSE), is able to extract these compo­ nents without much destructive effect on the background electroencephalogram (EEG). It is shown that the proposed method outperforms other methods particularly in pre­ serving the remaining signals. The CSE is utilized to remove the BCG artifact from real EEG data recorded inside the magnetic resonance (MR) scanner, i.e., visual evoked potential (VEP). The results are compared to the results of benchmark BCG artifact removal techniques. It is shown that VEPs recorded inside the scanner and processed using the proposed method are more correlated with the VEPs recorded outside the scanner. Moreover, there is no need for electrocardiogram (ECG) data in this method as the cycle frequency of the BCG artifact is directly computed from the contaminated EEG signals. To my beloved wife Saideh for all her support, patience and encouragement.