Spectro-Temporal Features for Robust Speech Recognition Using Power-Law Nonlinearity and Power-Bias Subtraction
Title | Spectro-Temporal Features for Robust Speech Recognition Using Power-Law Nonlinearity and Power-Bias Subtraction |
Publication Type | Conference Paper |
Year of Publication | 2013 |
Authors | Chang, S-Y., Meyer B. T., & Morgan N. |
Other Numbers | 3597 |
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. |
URL | https://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 |