Band-adaptive formant frequency estimation from noisy speech in correlation domain

Authors

  • Shaikh Anowarul Fattah Dept. of Electrical and Computer Engineering, Concordia University, 1455 de Maisonneuve Blvd. W., Montreal, QC H3G IM8, Canada
  • Wei-Ping Zhu Dept. of Electrical and Computer Engineering, Concordia University, 1455 de Maisonneuve Blvd. W., Montreal, QC H3G IM8, Canada
  • M. Omair Ahmad Dept. of Electrical and Computer Engineering, Concordia University, 1455 de Maisonneuve Blvd. W., Montreal, QC H3G IM8, Canada

Keywords:

Adaptive filters, Algorithms, Correlation detectors, Electric filters, Natural frequencies, Regression analysis, Autocorrelation functions, Correlation domains, Formant estimation, Formant frequency, Formant frequency estimation, Impact of noise, Natural speech, Noisy speech, Spectral peak

Abstract

A study was conducted to estimate formant frequencies accurately under a severe noisy condition. A band-limited repeated autocorrelation function (RACF) was introduced of the observed noisy speech to overcome the adverse impact of noise. It was shown that the RACF was pole-preserving and capable of reducing the effect of noise significantly. A band-adaptive filter-bank was employed on a zero lag compensated ACF to separate each formant frequency region before the formant estimation. Autocorrelation operation was repeated on each of the resulting band-limited ACFs and a spectral peak picking method was employed to each of the band-limited RACFs to extract formant frequencies. The proposed algorithm was tested on synthetic and natural speech signals in the presence of noise.

Published

2009-09-01

How to Cite

1.
Fattah SA, Zhu W-P, Ahmad MO. Band-adaptive formant frequency estimation from noisy speech in correlation domain. Canadian Acoustics [Internet]. 2009Sep.1 [cited 2021May13];37(3):92-3. Available from: https://jcaa.caa-aca.ca/index.php/jcaa/article/view/2149

Issue

Section

Proceedings of the Acoustics Week in Canada