摘要
提出了一种基于多级小波分解重构和非线性动力学参数的病理嗓音基音频率检测算法。首先对病理嗓音进行多级小波分解及重构,然后采用最大李雅普诺夫指数和近似熵表征不同重构嗓音的规则度从而自适应地选择周期性最优的小波重构嗓音信号,以直接提取基音频率。实验结果表明,与传统的基音检测算法相比,该方法有效地避免了检测中所存在的倍频及分频误差,提高了病理嗓音检测的鲁棒性及准确度。
This paper proposes a new pitch detection algorithm of pathological vome based on alscrete wave- let, largest lyapunov exponent and approximate entropy. Firstly, Daubechies＇ discrete wavelet transform is used to process the pathological voice, and then the most regular reconstructed signal is adaptively selected to detect period based on tile two nonlinear dynamic parameters. Results show that this method could effec- tively avoid the detection of frequency doubling and frequency division error, and improve the robustness and accuracy of the pathologic voice detection.
出处
《通信技术》
2013年第11期42-46,共5页
Communications Technology
基金
国家自然科学基金(No.61271359)病理嗓音识别与修复~~
关键词
病理嗓音基频
离散小波变换
最大李亚普诺夫指数
近似熵
pathological voice
pitch frequency
discrete wavelet transform
LLE （Largest Lyapunov Expo-nent）
ApEn （ Approximate Entropy）
