Non-linear Analysis Model Of Speech Signal Based On Neural Network And Its Application

Speech signal processing is one science with immense knowledge and extensive use; people have active demand for application. Speech signal processing technology has made great achievements, spectral analysis, wavelet analysis, vector quantization and dynamic time warping are formed, some successful speech signal analysis models have appeared, but there are still several problems in real application. Traditional speech signal analysis models are poor adaptability and simple in structure, have strong dependence on environment, so that the needs of modem speech signal processing can not be met.In this paper, the subject of this study is Chinese speech, and this paper focuses on the major problems in the procedure of Chinese speech signal processing. To improve the accuracy and solvability of speech signal analysis model, a homotopy analysis model based on neural network is proposed.Firstly, this paper does some research in classical speech signal linear analysis models and non-linear analysis models, compares and discusses the structure of these models, basic theory, the solution method and key technology in application, and emphasizes on homotopy model and relative theory. Then, on the basis of homotopy model, a method that uses neural network as a tool to solve the particular optimal non-linear function is given, and homotopy model based on neural network(HMNN) is proposed. This model is characterized by simple solution procedure, using neural network iterative method to solve the optimal non-linear function with specific form in homotopy model, and model parameters. Not only the complexity of computation is greatly reduced, but also the complicated non-linear optimization problem is avoided. Moreover, according to the expression and solution of HMNN model, this paper has discussed the problems about wavelet preconditioning and superiority rank of model. Finally, this paper has discussed the application in speech signal compression with HMNN model, and a speech signal compression coding algorithm based on HMNN model is proposed.The addition of studying structure and solution of HMNN model, applicable method of HMNN model is proposed, the feasibility and validity of HMNN model and speech signal compression coding algorithm based on HMNN model are verified through experiments and data.