Single Channel Blind Source Separation Method Research And Application

Blind Source Separation (BSS) is the separation of a set of source signalsfrom a set of mixed signals (observed signals), without the aid of informationabout the source signals or the mixing process. Single Channel Blind SourceSeparation (SCBSS) is an extreme case of underdetermined Blind SourceSeparation. In the real life, due to the restriction on the installation conditionsor costs, often encounter a Single sensor for multipath mixed signal to thecollected signals, and how to use the algorithm to estimate or restore themultiple source signals, is a very challenging problem in the field ofmathematic. This paper focuses on the single-channel blind source separationwhich is a hot part of signal processing at present, and analyses the singlechannel blind source separation algorithm. The main content of this paper areas follows:1. Put forward an adaptive single-input multiple-output blind sourceseparation.To resolve the poor adaptive degree and low iteration efficiency ofEEMD-ICA algorithm, EEMD-PCA-ICA algorithm is proposed in this paper.Finally compared it with the EEMD-ICA[29]and wavelet-ICA algorithm[26,27]in the recovery effect and running speed.2. Proposed a novel single-channel blind source separation method basedLMD algorithm.LMD algorithm is similar to the EMD algorithm that can according to thecharacteristics of signal, decompose it adaptively into components which havespecific physical meaning. Some papers have indicated that the LMDoutperformed EMD than the decomposition component and instantaneousfrequency. However, the algorithm also has end effect problem, end effect if notsuppressed, it will cause the algorithm to increase the number of iterations, theobtained component distortion high with iterative carried increasingly, seriouscan lead to get some meaningless component. In this paper, the adaptive phasesinusoidal continuation method was proposed to suppress LMD algorithm endpoint effect, then, based on the above optimized LMD algorithm, weproposed a new type of single-channel blind source separation. Finally, to provethe effectiveness of the present method, the actual signal was processed.