Research On Convolutiv Blind Signal Separation Algorithm Based On Biomimetic Intelligence

Blind signal separation algorithm is based on the statistical characteristics of the source signal,under the condition of unknown source and transmission channel parameters,and only uses the observation signals obtained by sensors to separate the independent source signals.And it has been applied in many fields.Linear mixture and convolutive mixture are the focuses of scholars.However,the convolutive mixture model is more consistent with the phenomenon of delay and attenuation when the sound propagates in the air,so the study of convolutive mixture is more meaningful.The method of solving the problem of convolutive mixed blind signal separation mainly consists of time domain method and frequency domain method.However,the time domain method involves complicated convolutive operations,which results in a lot of computation burden.Therefore,the frequency domain method is the key research object in this paper.The main contents of the research are as follows:(1)Proposed a inspired backtracking search algorithm and apply it to the problem of convolutive mixed blind signal separation.The traditional frequency domain method used in solving the problem of convolutive blind signal separation is usually the gradient algorithm or the Newton iterative algorithm.The step size,the nonlinear function and the initial matrix should be selected reasonably before solving the problem.Otherwise,the separation performance of the algorithm will be seriously affected.To solve the problem,a biomimetic intelligent algorithm with simple operation and excellent performance is used in this paper,that is,backtracking search algorithm.However,the backtracking search algorithm is difficult to balance in the two aspects of development and exploration.The paper improves backtracking search algorithm,that is,put forward a new search equation and add perturbation operator to the original search equation.The simulation experiment shows that the inspired backtracking search algorithm has better performance.Then,the inspired backtracking search algorithm is proposed to replace the traditional gradient optimization algorithm and optimize the separation matrix represented by the complex Givens matrix.Experiments show that the proposed algorithm achieves better separation performance than other algorithms.(2)Proposed a convolutive mixed blind signal separation algorithm based on complexvalued backtracking search.In order to solve the problem of many unknown parameters in the complex value separation matrix,the real particles in the backtracking search algorithm are extended into complex particles and form a solution space in a complex domain.Then the QR decomposition method is used to solve the separation matrix directly in the solution space of the complex field.Finally,the complex valued backtracking search algorithm is applied to optimize the unknown parameters in the separation matrix to find the global optimal solution.The simulation experiments show that the proposed algorithm reduces the parameters to be required,and improves the performance of the algorithm.(3)Proposed a convolutive mixed blind signal separation algorithm based on complex valued backtracking search and IVA.The traditional IVA algorithm is based on the gradient class optimization algorithm,which depends on the initial matrix and the step length.It is very easy to fall into the local optimal,when dealing with convolutive mixed blind signals.This paper proposes a method of optimizing the initial matrix of IVA algorithm by complex valued backtracking search algorithm.The mutual information of the estimated signals obtained from each frequency point is used as the objective function.The best solution searched by the backtracking search algorithm are used as the initial value of the separation matrix of the IVA algorithm.Finally,the IVA algorithm is used to solve and realize the separation.The simulation experiments show that the proposed algorithm not only eliminates the sort processing but also improves the separation performance.