语音信号非线性特征的数值表示

Numerical Description of the Nonlinear Aspect of Speech

分析了语音信号的非线性特征 ,用傅里叶谱、吸引子、分维数、李亚普诺夫指数等非线性动力学行为的数值特征表征了语音信号的非线性特征。采用时延语音序列重构语音吸引子 ,用计盒分维数计算语音波形的分维数。实验结果显示语音波形具有分形特征 。

The nonlinearity of speech is analyzed quantitatively in this paper.Four numerical characteristics of dynamics system,i.e.,Fourier spectrum,attractor,fractal dimension,and Lyapunov exponent,serve as the indicators of the nonlinearity of speech.In our study,time delayed samples of speech series are used to reconstruct speech attractor in phase space.The box counting fractal dimension of speech waveform is calculated to indicate the fractal aspect of speech.The study of the nonlinearity of speech concludes that speech waveform is fractal and that there exists chaos in speech.