Phase autocorrelation bark wavelet transform (PACWT) features for robust speech recognition

Sayf A. Majeed, Hafizah Husain, Salina Abdul Samad

Research output: Contribution to journalArticle

5 Citations (Scopus)

Abstract

In this paper, a new feature-extraction method is proposed to achieve robustness of speech recognition systems. This method combines the benefits of phase autocorrelation (PAC) with bark wavelet transform. PAC uses the angle to measure correlation instead of the traditional autocorrelation measure, whereas the bark wavelet transform is a special type of wavelet transform that is particularly designed for speech signals. The extracted features from this combined method are called phase autocorrelation bark wavelet transform (PACWT) features. The speech recognition performance of the PACWT features is evaluated and compared to the conventional feature extraction method mel frequency cepstrum coefficients (MFCC) using TI-Digits database under different types of noise and noise levels. This database has been divided into male and female data. The result shows that the word recognition rate using the PACWT features for noisy male data (white noise at 0 dB SNR) is 60%, whereas it is 41.35% for the MFCC features under identical conditions.

Original languageEnglish
Pages (from-to)25-31
Number of pages7
JournalArchives of Acoustics
Volume40
Issue number1
DOIs
Publication statusPublished - 2015

Fingerprint

speech recognition
wavelet analysis
autocorrelation
pattern recognition
digits
coefficients
white noise

Keywords

  • Feature extraction
  • Phase autocorrelation
  • Speech recognition
  • Wavelet transform

ASJC Scopus subject areas

  • Acoustics and Ultrasonics

Cite this

Phase autocorrelation bark wavelet transform (PACWT) features for robust speech recognition. / Majeed, Sayf A.; Husain, Hafizah; Abdul Samad, Salina.

In: Archives of Acoustics, Vol. 40, No. 1, 2015, p. 25-31.

Research output: Contribution to journalArticle

@article{6de2ee7c8d994c3eb5a60eee4087c2e2,
title = "Phase autocorrelation bark wavelet transform (PACWT) features for robust speech recognition",
abstract = "In this paper, a new feature-extraction method is proposed to achieve robustness of speech recognition systems. This method combines the benefits of phase autocorrelation (PAC) with bark wavelet transform. PAC uses the angle to measure correlation instead of the traditional autocorrelation measure, whereas the bark wavelet transform is a special type of wavelet transform that is particularly designed for speech signals. The extracted features from this combined method are called phase autocorrelation bark wavelet transform (PACWT) features. The speech recognition performance of the PACWT features is evaluated and compared to the conventional feature extraction method mel frequency cepstrum coefficients (MFCC) using TI-Digits database under different types of noise and noise levels. This database has been divided into male and female data. The result shows that the word recognition rate using the PACWT features for noisy male data (white noise at 0 dB SNR) is 60{\%}, whereas it is 41.35{\%} for the MFCC features under identical conditions.",
keywords = "Feature extraction, Phase autocorrelation, Speech recognition, Wavelet transform",
author = "Majeed, {Sayf A.} and Hafizah Husain and {Abdul Samad}, Salina",
year = "2015",
doi = "10.1515/aoa-2015-0004",
language = "English",
volume = "40",
pages = "25--31",
journal = "Archives of Acoustics",
issn = "0137-5075",
publisher = "de Gruyter",
number = "1",

}

TY - JOUR

T1 - Phase autocorrelation bark wavelet transform (PACWT) features for robust speech recognition

AU - Majeed, Sayf A.

AU - Husain, Hafizah

AU - Abdul Samad, Salina

PY - 2015

Y1 - 2015

N2 - In this paper, a new feature-extraction method is proposed to achieve robustness of speech recognition systems. This method combines the benefits of phase autocorrelation (PAC) with bark wavelet transform. PAC uses the angle to measure correlation instead of the traditional autocorrelation measure, whereas the bark wavelet transform is a special type of wavelet transform that is particularly designed for speech signals. The extracted features from this combined method are called phase autocorrelation bark wavelet transform (PACWT) features. The speech recognition performance of the PACWT features is evaluated and compared to the conventional feature extraction method mel frequency cepstrum coefficients (MFCC) using TI-Digits database under different types of noise and noise levels. This database has been divided into male and female data. The result shows that the word recognition rate using the PACWT features for noisy male data (white noise at 0 dB SNR) is 60%, whereas it is 41.35% for the MFCC features under identical conditions.

AB - In this paper, a new feature-extraction method is proposed to achieve robustness of speech recognition systems. This method combines the benefits of phase autocorrelation (PAC) with bark wavelet transform. PAC uses the angle to measure correlation instead of the traditional autocorrelation measure, whereas the bark wavelet transform is a special type of wavelet transform that is particularly designed for speech signals. The extracted features from this combined method are called phase autocorrelation bark wavelet transform (PACWT) features. The speech recognition performance of the PACWT features is evaluated and compared to the conventional feature extraction method mel frequency cepstrum coefficients (MFCC) using TI-Digits database under different types of noise and noise levels. This database has been divided into male and female data. The result shows that the word recognition rate using the PACWT features for noisy male data (white noise at 0 dB SNR) is 60%, whereas it is 41.35% for the MFCC features under identical conditions.

KW - Feature extraction

KW - Phase autocorrelation

KW - Speech recognition

KW - Wavelet transform

UR - http://www.scopus.com/inward/record.url?scp=84928242196&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84928242196&partnerID=8YFLogxK

U2 - 10.1515/aoa-2015-0004

DO - 10.1515/aoa-2015-0004

M3 - Article

AN - SCOPUS:84928242196

VL - 40

SP - 25

EP - 31

JO - Archives of Acoustics

JF - Archives of Acoustics

SN - 0137-5075

IS - 1

ER -