Research On Multi-target Separation And Location Based On Blind Signal Processing

Blind signal processing, a new signal processing technique during the past 2 decades, provide a brand new idea to multi-source signal processing. Based on the acoustic underwater weapons technology to the needs of practical application, this paper conducted a study of blind signal processing technology on acoustic signal separation and target localization. The major achievements are as follows:(1) This paper analysis contrastly several typical methods and realizable for Blind Signal Separation, and the relationship with independent component analysis (ICA). The key is derived based on the natural gradient instantaneous Blind Signal Separation(BSS) algorithm.(2) The article effectively improve the natural gradient: based on conventional NG, increase sub-sampling data function, use the rules of variable step, construct of a more fitting nonlinear signal source probability density function, and raise a performance indicators based on power spectrum. The improved algorithm can accelerate the convergence, strengthen the stability and enhance the adaptability.(3) From the perspective of information, the paper studies the linear blind deconvolution, successful separate two sources based on information maximization algorithm to the feedback network.(4) When signals are plus noise, the paper presents adaptive migration algorithm of two linear hybrid system. The experimental results show that the algorithm can achieve separation in a complex noise environment, and have stability.(5) Under the conditions of the element more than 3 in acoustic array, direction estimation method is proposed in two types blind separation of linear hybrid syst(?)m in this paper. Instantaneous blind separation complex algorithm can separate and locate acoustic multi-target signal mixed noise, and the position estimation error is less than 4%; Blind deconvolution algorithm can also separate acoustic twin targes signal.