Research On Single Channel Blind Source Separation Based On Random Signal

Blind Source Separation(BSS) is a process to reduct of the various source signals in the case of unknown signal sources and mixed-model, according to the received characteristics of the mixed signals. With the development of engineering and scientific research, many aspects have related to blind source separation technology, such as biomedical testing, speech signal recognition, mechanical fault diagnosis, wireless spectrum detection. Among them, single-channel blind source separation(SCBSS) is an extreme form of underdetermined blind source separation for the, which is very difficulty in the blind source separation using the observed single channel mixed signal to recover multiple source signals. This article focuses the the single-channel blind source separation of random signal on researching, mainly as follows:1. A new blind source separation method is put forward. The steps are as follows: firstly, multiple intrinsic mode function is disclosed from the received mixed-signal through ensemble empirical mode decomposition(EEMD). Secondly, an improved principal component analysis(PCA) is proposed to extract Principal component from multiple intrinsic mode function, which is to drop dimension again. Thirdly, independent components analysis(ICA) is used to complete the blind source separation of the random signals. Lastly, it is neccesary to compare the algorithm proposed by this paper with EEMD-PCA-ICA and EEMD-ICA to verify its efficiency through simulation.2. The single-channel blind source separation is carried out by using the "EEMD-PCA-ICA" and "EEMD-improvement PCA-ICA" for cycles-stationary random signal and non-stationary random signal. Experimental datas under the two methods are compared to research the differences and applicability in blind source separation for different random signals.3. The blind source separation of the convolved signals is researched. It is proposed for the indepenent vector analysis of variable step size gradient algorithm which is suitable for the Surface EMG signal decomposition. The IVA model is applied to the EMG of the convolutive mixtures to etract signal of mo-tor unit action potentials which implies in the Surface EMG signal. And then it is essential to analyze and compare performance between this algorithm and the method of independent component analysis for decomposition.