Abstract: A new method of the recurrence analysis pitch detection of nonlinear dynamical characteristics for speech signals is designed, which firstly calculates the pitch by recurrence quantification, and then distinguishes accurately voiced/unvoiced by the product of the recurrence degree and the pitch, and modifies the fluctuating pitch. The results show this method performance is better than the conventional autocorrelation algorithm and cepstrum method, especially in the part that the surd and the sonant are not evident, and can get a high robustness in noisy environment.

Key words: pitch detection, recurrence plot, recurrence analysis, feature extraction

摘要： 提出了一种语音信号非线性动力学特性递归分析的基音检测新方法；提出了递归数计算基音，研究了用递归度与所算基音之积精确区分清浊音法，给出了波动基音修正方法。实验结果表明，该方法可以得到比传统自相关法和倒谱法更好的结果，尤其是清浊音特征不明显的音段，基音检测的性能更好，在带噪语音音环境下也具有很好的适应性。

关键词: 基音检测, 递归图, 递归分析, 特征提取