The Study Of Blind Source Separation Algorithms Of The Noise Signal

In recent decades, Blind separation technology has never been interrupted as‘cocktail party’ effect occurs. As a new signal processing research methods, blind separation algorithms are closely watched by research workers of many areas. The so-called ‘blind’ is the source signal is mixed with an unknown signal together to form a mixed signal, and then the sensor for mixed-signal acquisition, the collected mixed signal is the signal through acquisition system analysis, the final separation from the unknown source signals mixed signal. With the era of continuous progress,more and more areas in the application of blind separation technology, for example biomedical, radar and sonar, image processing, seismic exploration, wireless communications. In practical applications, in various fields of applications for blind separation algorithm is at the ideal state of the signal separation. But there is an objective condition that cannot be ignored, namely the presence of the signal in the noise signal. The presence of noise signals not only change the performance of the algorithm lead to the error, but also to make some serious blind separation algorithm cannot correct operation, and then cannot separate the desire signal correctly.Therefore, blind signal separation algorithm with noise is an important object of study,this paper based on blind separation theory, focusing on the noisy sound signal blind separation algorithm.Firstly, a brief background and significance of research introduces blind separation algorithms, as well as the development of the status at home and abroad.And assumptions on blind source separation can be achieved and uncertainty analysis.Secondly, the mathematical model of blind separation, some definitions for blind separation algorithm, theorem, inference, related necessary and sufficient conditions on the feasibility and the principle of separation were analyzed. Make analysis and signal preprocessing methods by different forms of cost function, proposed three blind separation algorithm instantaneous mixture model-joint approximate diagonalization algorithm, Fast ICA algorithm and the largest information technology algorithms, as well as hybrid convolution mode. Through the analysis of instantaneous mixtureseparation and blind deconvolution algorithm mixed, noting the advantages and disadvantages of these algorithms, and propose two improved algorithms based on deficiencies, further simplifying the blind separation algorithm. Currently, these algorithms have been applied to among image processing, biomedical, wireless communications and other fields.Finally, depending on the application, via USB-1208 LS data acquisition card and Lab VIEW data acquisition program design and car engine sound signal as the research object, and the acquisition system and experimental test of blind separation algorithm applied a detailed description using adaptive step hybrid neural network blind separation algorithm for noisy car engine sound signal blind separation experiments, and to analyze the simulation results, realized the fault diagnosis of automobile engines. End of the article for a summary is presented prospects, but also pointed out the direction for further study.