Adaptive Voiced/Unvoiced Segmentation for Mandarin Syllable
-
摘要: 采用离散小波变换(DWT)实现汉语普通话音节的清/浊音分段,算法根据信号性质自适应地确定离散小波变换的尺度,具有较好的非特定人性质,并且对不同采样率及环境噪声有较强的适应性.测试了算法在男、女声,不同采样率及不同信噪比下的清/浊音分段算法的性能.在无噪情况下正确率为99.44%,在信噪比为30dB、15dB及5dB时正确率均可达99.20%,实验结果证明了算法的有效性和对噪声及非特定人的顽健性.Abstract: Discrete Wavelet Transformation(DWT) is used to realize voiced/unvoiced segmentation for Mandarin syllable.The algorithm is speaker-independent and robust for different sample rate and background noise, which defines the proper scale of DWT adaptively. Several experiments in different conditions,such as male /female,different sample rate and SNR, are given. It can achieve 99.44% for clean speech, and 99.20% for different SNR value. The effectiveness and robustness of the algorithm for noise and different speaker are proven by the experimental results.
-
Key words:
- discretization /
- digital signals /
- adaptive /
- wavelet transform /
- voiced/unvoiced segmentation /
- speaker-independent
-
[1] Burgnara F, Falavigna D, Omologo M. Automatic segmentation and labeling of speech based on hidden Markov models[J]. Speech Communication, 1993,12:357~370. [2] Qi Yingyong,Hunt Bobby R. Voiced-unvoiced-silence classification of speech using hybrid features and a network classifier[J]. IEEE Transactions on Speech and Audio Processing, 1993,1(2):250~255. [3] Medan Y, Yair E, Chazan D. Super resolution pitch determination of speech signals[J]. IEEE Transactions on Signal Processing, 1991,39(1):40~48. [4] Lin Ming-tzaw, Lee Ching-kuen, Lin Chin-yi. Consonant/vowel segmentation for Mandarin syllable recognition[J]. Computer Speech and Language,1999,13:207~222. [5] Kadambe S, Boudreaux-Bartels S F. Application of the wavelet transform for pitch detection of speech signals[J]. IEEE Transactions on Information Theory, 1992,38(2):917~924. -

计量
- 文章访问数: 4649
- HTML全文浏览量: 81
- PDF下载量: 1117
- 被引次数: 0