| Each of developing signal processing techniques could be used to medical signal processing soon, and make it great progress. The signals that take human function information are always mixed through a complex pattern. However, in order to analyze the physiology signal source and diagnose diseases, the clinicians desire almost acquire signal non-invasively and conveniently.Independent component analysis (ICA) is a new statistics signal processing method, developing with blind sources separation problem in 1990s. ICA can be used to decompose the mixed signal into independent components, and the independence of the decomposed component is emphasized. And now, independent component analysis is an attractive method used to extract the independent components such as in communications, speech recognition and bio-signal processing in recent years. In the medical signal processing, the multi-sources mixture problem and the mixed mode like as the ICA model are almost referred.All of the ICA applications to bio-medical signal processing are centralized in analyzing electroencephalogram components, extracting brain function information and someone else filtering the interference in EEG. Wavelet transform (WT) is a new branch in applied mathematics. Research fellows generally accepted that it is a perfect colligation of functional analysis, belt transect analysis, numerical analysis, harmonic analysis and Fourier analysis. It is the inheritance and great advancement for Fourier transform. In engineering application area, especially in nonlinear signal processing such as pattern recognition, quantum physics and communication it has a lot of novel applications. The mechanism of human sense system processing and thought pattern are seemed as wavelet analysis, so WT is suitable to extract bio-signal. The project will continue our research field in extracting deep information of physiology signals. It is important that ICA, WT and AIC will be employed to realize the extracting and analyzing methods for electrocardiogram and evoked potentials. The chief signal processing methods include WT, ICA and adaptive interference cancellation (AIC). The main research substance involves as follows:Firstly, ICA theories, deduction and some algorithms were addressed. It is also simple to introduce basic points of wavelet transform. One-dimension multi-resolution analysis (MRA) and continuous wavelet transform (CWT) had been explained selectively. Secondly, the electrocardiosignal model had been analyzed, and ICA was used to extract the independent components from 4-lead simultaneous recording ECS. Then, ICA and wavelet transform filtering noise had been combined to separate atrial fibrillation wave. We obtain the pure AF wave.Thirdly, we had used ICA and blind sources separation (BSS) to study the control pattern of ANS to heart, and then extract and analyze the sub-signals from HPS, which is controlled by ANS. The sub-signals could express the regulation of cardiac sympathetic and parasympathetic nerves. The physiology significances were confirmed through an experiment research that is a micro-sample about 10 examples.Fourthly, we designed an invasive experiment to acquire the rabbit cortical somatosensory evoked potential (SEP). The time-frequency feature was analyzed. Results have showed that variability of single-trail SEP latency becomes large with time in a stimulation period. Spectrum of SEP includes three main frequency spectrum packages (0~14Hz, 14-50Hz, above 50Hz). The technique of summation could result in a lot of signal aberration, such as waveform confluence, new waveform emerging and after-discharging components dismissing.Finally, Three modern signal-processing methods were employed to do single–trial evoked potential. One-dimension multi-resolution analysis (MRA) was used to decomposed original EPs and rebuilt a signal, which included evoked component with the frequency band ranged from 0 Hz to 60Hz. The rebuilt signal would be processed by continuous wavelet transform (CWT), which...
|
PDF Full Text Request