A new feature extraction front-end for robust speech recognition using progressive histogram equalization and multi-eigenvector temporal filtering

In this paper, feature extraction methods based on frequencywarped Minimum Variance Distortionless Response (MVDR) spectrum estimation are analyzed and tested. The effectiveness of the conventional FFT-based Mel-Frequency Cepstrum Coefficients (MFCCs) and the MVDR-based features are carefully compared. Two normalization techniques are further applied to improve the robustness of the features: the widely used cepstral normalization (CN), and newly proposed progressive histogram equalization (PHEQ). Extensive experiments with respect to the AURORA2 database were performed. The results indicated that both the MVDR-based features and the normalization processes are very helpful.

show abstract

Novel Feature Extraction for Noise Robust ASR using the Aurora 2 Database

Hix

Zahorian

Meng

2006 IEEE International Conference on Acoustics Speed and Signal Processing Proceedings

View full text Add to dashboard Cite

A new feature extraction front-end for robust speech recognition using progressive histogram equalization and multi-eigenvector temporal filtering

Cited by 5 publications

References 6 publications

Processing in the feature and model domains

Processing in the feature and model domains

Robust features for speech recognition using minimum variance distortionless response (MVDR) spectrum estimation and feature normalization techniques

Novel Feature Extraction for Noise Robust ASR using the Aurora 2 Database

Contact Info

Product

Resources

About