On the relative importance of various components of the modulation spectrum for automatic speech recognition

Noboru Kanedera, Takayuki Arai, Hynek Hermansky, Misha Pavel

Research output: Contribution to journalArticle

100 Citations (Scopus)

Abstract

We measured the accuracy of speech recognition as a function of band-pass filtering of the time trajectories of spectral envelopes. We examined (i) several types of recognizers such as dynamic time warping (DTW) and hidden Markov model (HMM), and (ii) several types of features, such as filter bank output, mel-frequency cepstral coefficients (MFCC), and perceptual linear predictive (PLP) coefficients. We used the resulting recognition data to determine the relative importance of information in different modulation spectral components of speech for automatic speech recognition. We concluded that: (1) most of the useful linguistic information is in modulation frequency components from the range between 1 and 16 Hz, with the dominant component at around 4 Hz; (2) in some realistic environments, the use of components from the range below 2 Hz or above 16 Hz can degrade the recognition accuracy.

Original languageEnglish (US)
Pages (from-to)43-55
Number of pages13
JournalSpeech Communication
Volume28
Issue number1
DOIs
StatePublished - May 1999

Fingerprint

Automatic Speech Recognition
Speech recognition
Modulation
Filter banks
Frequency modulation
Hidden Markov models
Linguistics
Trajectories
Dynamic Time Warping
Frequency Modulation
Filter Banks
bank
Coefficient
Speech Recognition
Range of data
Markov Model
Envelope
linguistics
Filtering
Trajectory

ASJC Scopus subject areas

  • Signal Processing
  • Electrical and Electronic Engineering
  • Experimental and Cognitive Psychology
  • Linguistics and Language

Cite this

On the relative importance of various components of the modulation spectrum for automatic speech recognition. / Kanedera, Noboru; Arai, Takayuki; Hermansky, Hynek; Pavel, Misha.

In: Speech Communication, Vol. 28, No. 1, 05.1999, p. 43-55.

Research output: Contribution to journalArticle

Kanedera, Noboru ; Arai, Takayuki ; Hermansky, Hynek ; Pavel, Misha. / On the relative importance of various components of the modulation spectrum for automatic speech recognition. In: Speech Communication. 1999 ; Vol. 28, No. 1. pp. 43-55.
@article{b3056bf3130849fcab0d91e8c41e8393,
title = "On the relative importance of various components of the modulation spectrum for automatic speech recognition",
abstract = "We measured the accuracy of speech recognition as a function of band-pass filtering of the time trajectories of spectral envelopes. We examined (i) several types of recognizers such as dynamic time warping (DTW) and hidden Markov model (HMM), and (ii) several types of features, such as filter bank output, mel-frequency cepstral coefficients (MFCC), and perceptual linear predictive (PLP) coefficients. We used the resulting recognition data to determine the relative importance of information in different modulation spectral components of speech for automatic speech recognition. We concluded that: (1) most of the useful linguistic information is in modulation frequency components from the range between 1 and 16 Hz, with the dominant component at around 4 Hz; (2) in some realistic environments, the use of components from the range below 2 Hz or above 16 Hz can degrade the recognition accuracy.",
author = "Noboru Kanedera and Takayuki Arai and Hynek Hermansky and Misha Pavel",
year = "1999",
month = "5",
doi = "10.1016/S0167-6393(99)00002-3",
language = "English (US)",
volume = "28",
pages = "43--55",
journal = "Speech Communication",
issn = "0167-6393",
publisher = "Elsevier",
number = "1",

}

TY - JOUR

T1 - On the relative importance of various components of the modulation spectrum for automatic speech recognition

AU - Kanedera, Noboru

AU - Arai, Takayuki

AU - Hermansky, Hynek

AU - Pavel, Misha

PY - 1999/5

Y1 - 1999/5

N2 - We measured the accuracy of speech recognition as a function of band-pass filtering of the time trajectories of spectral envelopes. We examined (i) several types of recognizers such as dynamic time warping (DTW) and hidden Markov model (HMM), and (ii) several types of features, such as filter bank output, mel-frequency cepstral coefficients (MFCC), and perceptual linear predictive (PLP) coefficients. We used the resulting recognition data to determine the relative importance of information in different modulation spectral components of speech for automatic speech recognition. We concluded that: (1) most of the useful linguistic information is in modulation frequency components from the range between 1 and 16 Hz, with the dominant component at around 4 Hz; (2) in some realistic environments, the use of components from the range below 2 Hz or above 16 Hz can degrade the recognition accuracy.

AB - We measured the accuracy of speech recognition as a function of band-pass filtering of the time trajectories of spectral envelopes. We examined (i) several types of recognizers such as dynamic time warping (DTW) and hidden Markov model (HMM), and (ii) several types of features, such as filter bank output, mel-frequency cepstral coefficients (MFCC), and perceptual linear predictive (PLP) coefficients. We used the resulting recognition data to determine the relative importance of information in different modulation spectral components of speech for automatic speech recognition. We concluded that: (1) most of the useful linguistic information is in modulation frequency components from the range between 1 and 16 Hz, with the dominant component at around 4 Hz; (2) in some realistic environments, the use of components from the range below 2 Hz or above 16 Hz can degrade the recognition accuracy.

UR - http://www.scopus.com/inward/record.url?scp=0032676337&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0032676337&partnerID=8YFLogxK

U2 - 10.1016/S0167-6393(99)00002-3

DO - 10.1016/S0167-6393(99)00002-3

M3 - Article

AN - SCOPUS:0032676337

VL - 28

SP - 43

EP - 55

JO - Speech Communication

JF - Speech Communication

SN - 0167-6393

IS - 1

ER -