Research And Application Of Single Channel Blind Signal Separation Based On EEMD

The technology of blind source separation have a potential application value in biomedical signal processing, speech signal processing, digital signal processing and other fields. It refers to the process that from the mixed signal(the observed signal) to isolate the source signal, under the condition of source signals and parameters of signal transmission channel are unknown. Existing blind source separation algorithms can solve the overdetermined or positive definite problem, in which the number of mixing signal is greater than the number of source signals or equal to the number of source signals. However, in the practical application, because of the equipment cost,installation conditions and other restrictions, usually can only rely on a sensor to receive the observation signal. Nevertheless, how to use the algorithm estimate and recovery source signal from mixing signal in a single channel, is a hot and difficult problem in the research field of blind source separation. The research of this article revolve around the single channel blind source separation(SCBSS), and applied it in reality signal processing, mainly do the following several aspects work:1)There are several types of the existing method of single channel blind source separation, including transform domain filtering, virtual multi-channel, base function and others. The method of virtual multi-channel is the foundation of this article, using the ensemble empirical mode decomposition(EEMD) make the single channel blind source separation problem transform to multi channel blind source separation problem;2)The detection and analysis technology of evoked potential(EP) signal is one of the important means in clinical diagnosis of neurological injury and disease. Existing methods can separate out EP signal from mixing signals after stimulation of hundreds times, but lost its instantaneous characteristics. Therefore, fewer times of extraction of EP signal is the focus of research in the current biomedical signal. This paper gives a SCBSS algorithm of EEMD-JADE, can isolated EP signal from the observed signals of single channel which mixing with EP and EEG signals of only 2 cycles. The simulation result proves the validity of EEMD-JADE algorithm. And compared with the EEMD-PCA-ICA method, the experimental results show that EEMD-JADE algorithm has better performance in isolating EP signal from observed signals of single channel of only 2 cycles;3)Speech is an important part of human life, but expect speech which received from the nature often mixed with environmental noise and other interference sources. On the basis of the existing algorithm, this paper gives an SCBSS algorithm based on EEMD-Sparse Representation, can isolated source signal of speech from the observed signals of single channel which random mixing with two kinds of speech signals.The simulation result proves the validity of EEMD-Sparse Representation algorithm. And compared with the EEMD-PCA-ICA method, the experimental results show that EEMD-Sparse Representation algorithm has better performance in isolating mixed speech signal of single channel;...