Autoregressive Hidden Markov Model and the Speech Signal

Jacob D. Bryan, Stephen E. Levinson

Research output: Contribution to journalConference article

Abstract

This paper introduces an autoregressive hidden Markov model (HMM) and demonstrates its application to the speech signal. In this variant of the HMM the observed signal is assumed to be Gaussian autoregressive and the probability density function is derived based on an approximation of the linear prediction error. A Baum-Welch style set of re-estimation formulas are then derived and used to infer the model parameters for a given data set, which correspond to linguistic structure in the context of speech data. The new set of re-estimation formulas are then applied to speech data and experimental results demonstrate inference of broad phonetic categories without prior knowledge of linguistic information. The experimental results and stability of this model are then briefly contrasted with historic experiments wherein phonetic information has been inferred directly from the speech signal using a similar autoregressive model.

Original languageEnglish (US)
Pages (from-to)328-333
Number of pages6
JournalProcedia Computer Science
Volume61
DOIs
StatePublished - Jan 1 2015
EventComplex Adaptive Systems, 2015 - San Jose, United States
Duration: Nov 2 2015Nov 4 2015

Fingerprint

Hidden Markov models
Speech analysis
Linguistics
Probability density function
Experiments

Keywords

  • autoregressive HMM
  • hidden Markov model
  • linear prediction
  • speech signal processing

ASJC Scopus subject areas

  • Computer Science(all)

Cite this

Autoregressive Hidden Markov Model and the Speech Signal. / Bryan, Jacob D.; Levinson, Stephen E.

In: Procedia Computer Science, Vol. 61, 01.01.2015, p. 328-333.

Research output: Contribution to journalConference article

@article{4d0e185db4814747877f170188babfc2,
title = "Autoregressive Hidden Markov Model and the Speech Signal",
abstract = "This paper introduces an autoregressive hidden Markov model (HMM) and demonstrates its application to the speech signal. In this variant of the HMM the observed signal is assumed to be Gaussian autoregressive and the probability density function is derived based on an approximation of the linear prediction error. A Baum-Welch style set of re-estimation formulas are then derived and used to infer the model parameters for a given data set, which correspond to linguistic structure in the context of speech data. The new set of re-estimation formulas are then applied to speech data and experimental results demonstrate inference of broad phonetic categories without prior knowledge of linguistic information. The experimental results and stability of this model are then briefly contrasted with historic experiments wherein phonetic information has been inferred directly from the speech signal using a similar autoregressive model.",
keywords = "autoregressive HMM, hidden Markov model, linear prediction, speech signal processing",
author = "Bryan, {Jacob D.} and Levinson, {Stephen E.}",
year = "2015",
month = "1",
day = "1",
doi = "10.1016/j.procs.2015.09.151",
language = "English (US)",
volume = "61",
pages = "328--333",
journal = "Procedia Computer Science",
issn = "1877-0509",
publisher = "Elsevier BV",

}

TY - JOUR

T1 - Autoregressive Hidden Markov Model and the Speech Signal

AU - Bryan, Jacob D.

AU - Levinson, Stephen E.

PY - 2015/1/1

Y1 - 2015/1/1

N2 - This paper introduces an autoregressive hidden Markov model (HMM) and demonstrates its application to the speech signal. In this variant of the HMM the observed signal is assumed to be Gaussian autoregressive and the probability density function is derived based on an approximation of the linear prediction error. A Baum-Welch style set of re-estimation formulas are then derived and used to infer the model parameters for a given data set, which correspond to linguistic structure in the context of speech data. The new set of re-estimation formulas are then applied to speech data and experimental results demonstrate inference of broad phonetic categories without prior knowledge of linguistic information. The experimental results and stability of this model are then briefly contrasted with historic experiments wherein phonetic information has been inferred directly from the speech signal using a similar autoregressive model.

AB - This paper introduces an autoregressive hidden Markov model (HMM) and demonstrates its application to the speech signal. In this variant of the HMM the observed signal is assumed to be Gaussian autoregressive and the probability density function is derived based on an approximation of the linear prediction error. A Baum-Welch style set of re-estimation formulas are then derived and used to infer the model parameters for a given data set, which correspond to linguistic structure in the context of speech data. The new set of re-estimation formulas are then applied to speech data and experimental results demonstrate inference of broad phonetic categories without prior knowledge of linguistic information. The experimental results and stability of this model are then briefly contrasted with historic experiments wherein phonetic information has been inferred directly from the speech signal using a similar autoregressive model.

KW - autoregressive HMM

KW - hidden Markov model

KW - linear prediction

KW - speech signal processing

UR - http://www.scopus.com/inward/record.url?scp=84962681661&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84962681661&partnerID=8YFLogxK

U2 - 10.1016/j.procs.2015.09.151

DO - 10.1016/j.procs.2015.09.151

M3 - Conference article

AN - SCOPUS:84962681661

VL - 61

SP - 328

EP - 333

JO - Procedia Computer Science

JF - Procedia Computer Science

SN - 1877-0509

ER -