Call for Paper - November 2022 Edition
IJCA solicits original research papers for the November 2022 Edition. Last date of manuscript submission is October 20, 2022. Read More

Noisy Speech Recognition by Mel-LPC based AR-HMM with Power and Time Derivative Parameters

Print
PDF
International Journal of Computer Applications
Foundation of Computer Science (FCS), NY, USA
Year of Publication: 2018
Authors:
M. Babul Islam
10.5120/ijca2018917149

Babul M Islam. Noisy Speech Recognition by Mel-LPC based AR-HMM with Power and Time Derivative Parameters. International Journal of Computer Applications 180(42):1-5, May 2018. BibTeX

@article{10.5120/ijca2018917149,
	author = {M. Babul Islam},
	title = {Noisy Speech Recognition by Mel-LPC based AR-HMM with Power and Time Derivative Parameters},
	journal = {International Journal of Computer Applications},
	issue_date = {May 2018},
	volume = {180},
	number = {42},
	month = {May},
	year = {2018},
	issn = {0975-8887},
	pages = {1-5},
	numpages = {5},
	url = {http://www.ijcaonline.org/archives/volume180/number42/29408-2018917149},
	doi = {10.5120/ijca2018917149},
	publisher = {Foundation of Computer Science (FCS), NY, USA},
	address = {New York, USA}
}

Abstract

In this paper, AR-HMM on mel-scale with power and Mel-LPC based time derivative parameters has been presented for noisy speech recognition. The mel-scaled AR coefficients and melprediction coefficients for Mel-LPC have been calculated on the linear frequency scale from the speech signal without applying bilinear transformation. This has been done by using a first-order allpass filter instead of unit delay. In addition, Mel-Wiener filter has been applied to the system to improve the recognition accuracy in presence of additive noise. The proposed system is evaluated on Aurora 2 database, and the overall recognition accuracy has been found to be 80.02% on the average.

References

  1. Juang, B. and Rabiner, L. R. 1986. Mixture autoregressive hidden Markov models for speech signals. IEEE Trans. Acoust., Speech, Signal Processing, 33: 1404-1413.
  2. Ephraim, Y. 1992. Gain adapted hidden Markov models for recognition of clean and noisy speech. IEEE Trans. Signal Processing, 40(6): 1303-1316.
  3. Ruske, G. and Lee, K. Y. 1999. Speech recognition and enhancement by a nonstationary AR HMM with gain adaptation under unknown noise. Proceedings ICASSP’99.
  4. Deng, L. 1992. A generalized hidden Markov model with state conditioned trend functions of time for speech signal. Signal Processing, 27: 65-72.
  5. Lee, K. Y. and Lee, J. 2001. Recognition of noisy speech by a nonstationary AR HMM with gain adaptation under unknown noise. IEEE Trans. Speech and Audio Processing, l 9(7): 741- 746.
  6. Logan, B. T. and Robinson, A. J. 1997. Improving autoregressive hidden Markov model recognition accuracy using a nonlinear frequency scale with application to speech enhancement. Proc. of EUROSPEECH, 2103-2106.
  7. Juang, B. 1984. On the hidden Markov model and dynamic time warping for speech recognition - a unified view. AT&T Bell Lab. Tec. Journal, 63(7): 1213-1243.
  8. Strube, H. W. 1980. Linear prediction on a warped frequency scale. J. Acoust. Soc. America, 68(4): 1071-1076.
  9. Matsumoto, H., et al. 1998. An efficient Mel- LPC analysis method for speech recognition. Proc. of ICSLP98: 1051- 1054.
  10. Islam, M. B., et al. 2007. Mel-Wiener filter for Mel-LPC based speech recognition. IEICE Transactions on Information and Systems, E90-D (6): 935-942.
  11. Oppenheim, A. V. and Johnson, D. H. 1972. Discrete representation of signals. IEEE Proc., 60(6): 681-691.
  12. Hirsch, H. G. and Pearce, D. 2000. The AURORA experimental framework for the performance evaluation of speech recognition systems under noisy conditions. Proc. ISCA ITRW ASR2000: 181-188.
  13. Leonard, R. G. 1984. A database for speaker independent digit recognition. ICASSP84, 3: 42.11.1-42.11.4.

Keywords

AR-HMM, Mel-LPC, Mel-Wiener filter, Aurora 2 database