噪声环境下语音分形特征的提取和分析

包永强; 赵力; 邹采荣

doi:10.3724/SP.J.1146.2005.00836

噪声环境下语音分形特征的提取和分析

doi: 10.3724/SP.J.1146.2005.00836

基金项目:

教育部基金(03082)和国家自然科学基金(60472058)资助课题

计量
- 文章访问数: 3412
- HTML全文浏览量: 98
- PDF下载量: 752
- 被引次数: 0
出版历程
- 收稿日期: 2005-07-13
- 修回日期: 2006-01-02
- 刊出日期: 2007-03-19

The Abstraction and Analysis of Fractal Characteristic of Noisy Speech

摘要

摘要: 该文针对目前的分维计算方法盒维、关联维等精度虽高，但计算复杂，Katz维计算简单、抗噪性能好、但精度不高的现状，提出了一种改进的基于波形的算法IBW-FD，分析了对分形布朗曲线、含噪语音(高斯白噪声，三种非平稳噪声)的性能。理论分析和实验结果表明： IBW-FD算法具有更强区分高斯白噪声和语音信号的能力；IBW-FD算法抗平稳和非平稳噪声能力要普遍好于盒维和Katz维。结果表明IBW-FD算法在复杂度、精确度和抗噪性能方面均优于现有的分维算法，是一种比较好的分维计算方法，不仅可以应用在语音处理中，而且也可应用于其它信号处理中。
- 语音信号处理;分形维;语言增强
Abstract: According to simple computation, good anti-noise ability and low precision of Katz algorithm and complex computation and good precision of box-counting dimension and correlation dimension, an Improvement fractal algorithm Based on Wave (IBW) is presented and analyzed through the fractal Brown curve and noisy speech according to the characteristic of the box dimension and Katz dimension. The theory analyse and experiment show that IBW-FD has lower computation and higher precision than Katz dimension and box-counting dimension. IBW-FD also has stronger ability of anti-noise and distinguish Gaussian noise and speech than the others. It shows that IBW-FD is the good speech fractal algorithm because of low complexity, good precision and nice anti-noise ability.

HTML全文

参考文献(1)

[1] Berry D A, Herzel H, Titze I R, and Krischer K. Interpretation of biomechanical simulations of normal and chaotic vocal fold oscillations with empirical eigenfuctions. J. Acoust. Soc. Am, 1994, 95(6): 3594-3604. [2] Kumar A and Mullick S K. Nonlinear dynamical analysis of speech[J].J. Acoust. Soc. Am.1996, 100(1):615-620 [3] Mandelbrot B B and Vess J W V. Fractional Brownian motion, fractional noise and application. SIAM Review, 1968, 10(4): 540-546. [4] Boshoff H F V. A fast box counting algorithm for determining the fractal dimension of sampled continuous functions. COMSIG92., Proceedings of the 1992 South African Symposium on 11 Sept., 1992: 43-48. [5] Pitsikalis V and Maragos P. Speech analysis and feature extraction using chaotic models[J].ICASSP2002, IEEE International Conference on, Orlando FL.2002, 1:533- [6] Gu Lingyun.[J].Gao Jianbo, and Harris G J. Endpoint detection in noisy environment using a pointcare recurrence metric. ICASSP0.2003,:- Yang Su and Li Zong-ge. A fractal based voice activity detector for internet telephone, ICASSP 2003, IEEE International Conference on, Hongkong, 2003: I808-I811. [7] Seo Jungpa, Hong S, Kim M, Baek I, Kwon Y, and Lee K. New speaker recognition feature using correlation dimension. ISIE 2001, IEEE International Symposium on Pusan, 2001, 1: 505-507. [8] Petry A and Barone D A C. Fractal dimension to speaker identification[J].ICASSP01, 2001 IEEE International Conference on Salt Lake City, UT.2001, 1:405-408 [9] Katz M. Fractals and the analysis of waveforms[J].Comput. Biol. Med.1988, 18(3):145-156 [10] Esteller R and Vachtsevanos G. A comparison of waveform fractal dimension algorithms. IEEE Trans. on Circuits and Systems, 2001, 48(2): 177-183.

施引文献

资源附件(0)

访问统计