基金项目:吉林省科技发展计划项目(201101110)；吉林市科技发展项目(2013625009)Project Supported by the Scientific Research Fundation of the Education Department of Jilin Province (No.201101110) and the Scientific Research Fundation of the Education Department of Jilin City (No.2013625009)
This paper presents an improved mutual information minimization algorithm for nonlinear blind source separation,perfects the optimization algorithm in terms of the crosstalk error(ECT) big,etc.The method uses the natural gradient optimization algorithm to optimize the objective function,this can avoid the matrix inversion calculation and reduce the computing time.Focus is on the introduction of disturbance signals in the network to optimize the process parameters.It prevents stagnation in the late algorithm,improves the ability of the optimization of nonlinear blind source separation algorithm.Experimental results show that the proposed nonlinear blind source separation optimization algorithm is a good algorithm,obtains effective,stable separation results.The experiment proves that the new optimization algorithm is superior to the traditional optimization algorithm.
