Regularized minimum variance distortionless response-based cepstral features for robust continuous speech recognition

@article{Alam2015RegularizedMV,
  title={Regularized minimum variance distortionless response-based cepstral features for robust continuous speech recognition},
  author={Md. Jahangir Alam and Patrick Kenny and Douglas D. O'Shaughnessy},
  journal={Speech Communication},
  year={2015},
  volume={73},
  pages={28-46}
}
In this paper, we present robust feature extractors that incorporate a regularized minimum variance distortionless response (RMVDR) spectrum e stimator instead of the discrete Fourier transform-based direct spectrum estimator, used in many front-ends including the conventional MFCC, to estimate the speech power spe ctrum. Direct spectrum estimators, e.g., single tapered periodogram, have high varianc e d they perform poorly under noisy and adverse conditions. To reduce this performance drop we… CONTINUE READING