基于混沌理论的汉语语音信号的分析与预测

The analysis and prediction of Chinese speech signal based on chaos theory

目前语音信号的分析与预测都是采用线性理论和线性预测技术,而语音信号的产生系统是一个复杂的非线性时变系统,而且具有混沌性,所以采用线性方法是不够的。深入研究了汉语语音信号的非线性特性,包括相空间重构理论及延迟时间、嵌入维数等相空间重构参数的确定方法,并求解出汉语语音音素的李雅普诺夫指数、延迟时间、嵌入维数和关联维数,所得结果表明汉语语音信号既非确定性的信号,又非随机信号,而是具有混沌特性的信号;根据汉语语音音素的延迟时间及嵌入维数的均值确定RBF神经网络(Radical Basis Function Network)模型中三层网络的神经元个数,结合RBF神经网络分析方法构造了一个非线性预测模型。仿真结果表明:基于RBF神经网络构造的非线性预测模型与线性预测模型相比,预测误差明显减小,预测性能上有所提高。

At present, the analysis and prediction of speech signal are all using linear theory and linear prediction technique, but the speech production system is complicated nonlinear and chaotic system, so linear methods are inadequate. Chaos theory is used in studying the nonlinear characteristic of Chinese speech, including theory of phase space reconstruction and methods of solving phase space reconstruction parameters containing delay time, embed dimension and so on, and solved the Lyapunov component, delay timemhed dimension and correlation dimension of Chinese speech phonemes, which indicate Chinese speech is not deterministic signal and stochastic signal, but has chaotic characteristics; the averages of the delay time and embedding dimension for Chinese speech phonemes determine the neurons number of three-layer network for Radical Basis Function （RBF） neural network model, RBF network analysis methods are applied to construct nonlinear predictor. Tile simulation results indicate： compared with the linear predictor, prediction error of nonlinear predictor based on RBF network is significantly decreased and has higher performance.