Study On The Methods Of Extracting Weak Biomedical Signal And Applications

Biomedical signals are always buried into the strong noise. Therefore, it's very important to remove noise and extract feature signal for biomedical science research and medical diagnosis. Based on the knowledge of present methods used to process biomedical signals and the characteristics of mixed signals, we remove the noise and extract feature signal effectively by choosing Wavelet Transform(WT) and Independent Component Analysis(ICA) algorithms.It reviews the traditional signal processing methods used in biomedical domain firstly. Then it discusses the theory of various time-frequency analysis methods and blind source separation methods which are focused on WT and ICA algorithms. In the paper, it compares the performance of the Infomax and Extended Infomax algorithms. Some experiments are performed on recorded data to verify the effective of the WT on the removing of power line noise in EEG and baseline noise in ECG; some experiments indicate that ICA algorithms show excellent performance on removing noise and extracting feature signal. In the paper, it also studies on the synthetic methods, WT combining with adding and ICA combining with adding, and then does some experiments to extract VEP from the recorded data based on synthetic methods. In the experiment based on the method of ICA combining with adding, it introduces the reference and discusses how the reference affects the performance of the separation.