Speech Signal Enhancement Research Based On Blind Source Separation

In the process of the actual speech signal transmission, the source signals are interfered by ambient noise signal, so that it affects the quality of the source signals,which affect communication effect. Therefore, it is necessary to enhance speech signal with noise and effectively inhibit the interference of environmental noise, so as to improve the quality of speech signal communication.Based on speech signal as the research object, and the paper based on blind source separation algorithm, after a lot of reading related literature both at home and abroad, I use fruit fly optimization algorithm to enhance the speech source signals.The simulation of speech signal enhancement algorithm and the results are studied and discussed by means of computer simulation technology.The paper adopts the blind source signal processing method. The selection of optimization algorithm is one of the core content.Considering that in the fruit fly optimization algorithm with parameter optimization is simple and easy to implement, and the advantages of high precision.I decided to adopt fruit fly optimization algorithm after learning all kinds of optimization algorithm research.The principle and application of the algorithm are expounded in this paper. Its improvement methods are also illustrated.Because the stability of fruit fly algorithm is poorer so as to easily plunge into local optimum, so the paper improve the traditional algorithm by changing the parameters of itself.The algorithm is improved through the optimization process,which is divided into two stages in the paper, the search step value increases gradually in the first stage, in the second stage it gradually decreases, change the definition of concentration determination value in the original fruit fly optimization algorithm and so on,so as to increase the diversity of the group.At the same time reduces the likelihood of trapping in local optimum, then to improve the convergence precision and convergence speed.In addition, Copy it and carry on the disturbance after getting the optimal solution. According to certain probability for the gaussian mutation processing,second optimization is carried out for individual mutated flies so as to avoid falling into local optimum.The traditional algorithm is improved by adopting the combination of optimized methods.Finally the simulation experiment research is carried out through improved fruit fly optimization algorithm for speech signal enhancement. Successive extract of the optimal solution and compare the results through speech signal quality evaluationindex,so as to obtain the experimental conclusion that is successfully obtain source signal from the mixed signals.