Research On Speech Signal Prediction Model Based On GP Algorithm

With the further development of information technology, higher speed and quality requirement of speech signal transmission is putting forward, and the traditional method of the speech signal coding in this period did not achieve greater development in theory. It is found in the speech signal nonlinear characteristic research that the speech signal and its time series both exhibit very complex nonlinear process. In this paper, a chaotic time series bidirectional prediction model of a series speech signal is built by improved GP algorithm based on the speech signal’s chaotic characteristics which enables the speech signal coding.(1)Collect the speech signal simple, and analyze the chaos characteristics of speech signal. The result indicates that speech signals have chaos characteristics and nonlinear method can be used to predict chaotic signal.(2)Improve the genetic programming algorithm in order to make the genetic algorithm more effectively in establishing speech signal nonlinear model in three aspects. Firstly, change the question’s description of hierarchy into the fix length linear structure to avoid the "Scale explosion". Secondly, we adopt multi-population genetic programming algorithm to increase the diversity of solution and improve the global search ability. Finally, the mountain climbing algorithm is introduced to further optimize the coefficients of the model.(3) Indicate the bidirectional prediction method based on the chaos theory, and describe the process of establishing the forecast model of speech signal using improved genetic programming algorithm. After preprocessing speech signal and reconstruction of phase space, establish the forecast model of each frame using the improved GP algorithm. Then by analyzing the structure of each model, one or a set of standardized model with certain generalization ability can be chosen. At last, each frame of speech data’s coefficients is determined by using optimization algorithms, to realize the speech signal coding.