Research Of Dynamic Blind Source Separation And Its Application To Biomedical Signal Processing

Blind source separation (BSS) is a multidimensional signal processing mode, which is developed in recent years. Its characteristic is that recovery unknown source signals only by observed signals without knowing the channel information. According to this characteristic, BSS has developed rapidly, and it has been used in many fields such as wireless communication, biomedical signal processing, and audio signal processing, digital image processing and pattern recognition. Especially in the field of biomedical signal processing, using blind source separation technique to extract the biomedical signals, which are submerged by interference and noise, become a hotspot research topic in signal processing.This thesis mainly study a new online blind source separation algorithm——the dynamic blind source separation algorithm based on sliding window, and applies it to detect the pulse signal of blood oxygen signal and separate fetal electrocardiogram(FECG) in biomedical signal processing fields. The concrete content includes the following respects:1. Study the dynamic blind source separation algorithm based on sliding window, analysis the factors, which affect the arithmetic performance, through the simulation, and propose that it don't need to use a variable step-length to improved the performance of algorithm, because the algorithm based on calculating block of data to update the separable matrix, cause the convergence of the algorithm is good, but compared with the online algorithm by single sample updated, its computation is enormous, so the recursive form of the algorithm is given.2. Through analyzing the principles of blood oxygen saturation measurement, it is demonstrated that extraction of the clean pulse signals of blood oxygen is key in testing SO2, and the dynamic blind source separation algorithm based on sliding window is introduced to remove the noise of the pulse signals. Four groups of data which are collected in different environment, another is one group data which is collected when the state of the tester is changing. The results show that, the algorithm can real-timely remove noise interference by tracking the state of test environment and the tester; it has good application value in testing SO2 process.3. In view of some problem in the current FECG extraction process, an online FECG separation algorithm is proposed. The experimental results indicate that the performance of this algorithm is better than the performance of usual adaptive blind source separation algorithm. The result of processing the real record FECG signal further validates the effectiveness of the algorithm, and also shows that the algorithm has wide application prospects.