Research of Robust Feature for Speech Recognition

Abstract:

Feature extraction plays an important role in speech recognition. In this paper, we propose a speech feature extraction scheme which focuses on increasing the robustness of speech recognizer in noise (additive) and channel (convolutive) distortion environment. Considering the two distortions are additive in spectral and log-spectral domain, respectively, we remove the additive components by computing the time derivatives of speech frames firstly in spectral domain and then in log-spectral domain. Compared with conventional methods, this method does not need spectrum estimation and prior knowledge of noise. Experimental results confirm that our proposed method can improve the speech recognition performance in environ-ments existing both noise and channel distortions.

• Periodical: Advanced Materials Research (Volumes 532-533)
• Pages: 1162-1166
• DOI: https://doi.org/10.4028/www.scientific.net/AMR.532-533.1162
• Online since: June 2012
• Authors: Xiang Hua Ren, Yun Xia Jiang
• Keywords: Channel Distortion, Feature Enhancement, Feature Extraction, Noise Distortion, Robust Speech Recognition

References

[1] J. Ortega-Garcia and L. Gonzalez-Rodriguez, in Overview of Speaker Enhancement Techniques For Automatic Speaker Recognition, edited by Proceeding of ICSLP, Philadelpia, PA( 1996).

DOI: 10.1109/icslp.1996.607754

[2] Acero, A.,. in: Acoustical and Environmental Robustness in Automatic Speech Recognition, edited by Kluwer Academic Publishers, (1993).

[3] Li, J., Deng, L., Yu, D., Gong, Y., Acero, in: High-performance HMM Adaptation with Joint Compensation of Additive and Convolutive Distortions Via Vector Taylor Series, edited by Proceeding of ASRU (2007).

DOI: 10.1109/asru.2007.4430085

[4] Saeed V., in : Advanced Digital Signal Processing and Noise Reduction, edited by John Wiley & Sons Inc(2005).

[5] Information on http: /www. speechocean. com/introduction/King-ASR-009. pdf.