Bounded cepstral marginalization of missing data for robust speech recognition

@article{Kafoori2016BoundedCM,
  title={Bounded cepstral marginalization of missing data for robust speech recognition},
  author={Kian Ebrahim Kafoori and Seyed Mohammad Ahadi},
  journal={Computer Speech & Language},
  year={2016},
  volume={36},
  pages={1-23}
}
Robust recognition of noisy speech achieved via a novel missing data technique.Proposed modified bounded marginalization compatible with MFCC trained models.The second proposed technique is more accurate, but still fast and simple.The third method competes with imputation techniques considering accuracy.Proposed techniques are all simpler and faster than imputation techniques. Spectral imputation and classifier modification can be counted as the two main missing data approaches for robust… CONTINUE READING

Figures, Tables, and Topics from this paper.

References

Publications referenced by this paper.
SHOWING 1-10 OF 42 REFERENCES

A waveletbased thresholding approach to reconstructing unreliable spectrogram components

B. adiezadegan, R. C. Rose
  • Speech Commun .
  • 2015

A Direct Masking Approach to Robust ASR

  • IEEE Transactions on Audio, Speech, and Language Processing
  • 2013

Coupling binary masking and robust ASR

  • 2013 IEEE International Conference on Acoustics, Speech and Signal Processing
  • 2013

Combining Speech Fragment Decoding and Adaptive Noise Floor Modeling

  • IEEE Transactions on Audio, Speech, and Language Processing
  • 2012
VIEW 1 EXCERPT

Combining missingdata reconstruction and uncertainty decoding for robust speech recognition

J. A. onzález, A. M. Peinado, N. Ma, A. M. Gómez, J. Barker
  • 2012
VIEW 1 EXCERPT