Spectro-Temporal Features for Robust Speech Recognition Using Power-Law Nonlinearity and Power-Bias Subtraction

TitleSpectro-Temporal Features for Robust Speech Recognition Using Power-Law Nonlinearity and Power-Bias Subtraction
Publication TypeConference Paper
Year of Publication2013
AuthorsChang, S-Y., Meyer B. T., & Morgan N.
Other Numbers3597
Abstract

Previous work has demonstrated that spectro-temporal Gabor features reduced word error rates for automatic speech recognition under noisy conditions. However, the features based on melspectra were easily corrupt ed in the presence of noise or channel distortion. We have exploited an algorithm for power normalized cepstral coefficients (PNCCs) to generate a more robust spectro-temporal representation. We refer to it as power normalized spectrum (PNS), and to the corresponding output processed by Gabor filters and MLP nonlinear weighting as PNS-Gabor. We show that the proposed feature outperforms state-

Acknowledgment

This work was partially supported by funding provided to ICSI by the U.S. Defense Advanced Research Projects Agency (DARPA) under contract number D10PC20024. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors or originators and do not necessarily reflect the views of DARPA or of the U.S. Government.

URLhttps://www.icsi.berkeley.edu/pubs/speech/spectrotemporal13.pdf
Bibliographic Notes

Proceedings of the IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2013), Vancouver, Canada

Abbreviated Authors

S.-Y. Chang, B. Meyer, and N. Morgan

ICSI Research Group

Speech

ICSI Publication Type

Article in conference proceedings