Estimation of syntactic structure using prosodic parameters and its application to speech recognition

Takashi Endou, Mamoru Watado, Tetsunori Kobayashi, Katsuhiko Shirai

Research output: Chapter in Book/Report/Conference proceedingChapter

Abstract

Prosodic information is very useful for the estimation of syntactic structure. So Phrase-level segmentation of continuous speech using pitch and power information is discussed to utilize for the word spotting. First, a rule based approach using pitch and power pattern is carried out. Then, a statistical approach is carried out, where the vector sequence of pitch and power are matrix quantized and conditional probabilities of the codes and the labels are calculated. The experimental result of the rule-based approach shows that the boundary detection accuracy is 72.8% in case that the adjacent phrases have loose relation, and 43.4% in case of tight relation. As for the statistical method, the boundary detection accuracy is 63.0%.

Original languageEnglish
Title of host publicationBulletin of Centre for Informatics, Waseda University
Pages17-23
Number of pages7
Volume15
Publication statusPublished - 1992 Jun
Externally publishedYes

Fingerprint

Syntactics
Speech recognition
Labels
Statistical methods

ASJC Scopus subject areas

  • Engineering(all)

Cite this

Endou, T., Watado, M., Kobayashi, T., & Shirai, K. (1992). Estimation of syntactic structure using prosodic parameters and its application to speech recognition. In Bulletin of Centre for Informatics, Waseda University (Vol. 15, pp. 17-23)

Estimation of syntactic structure using prosodic parameters and its application to speech recognition. / Endou, Takashi; Watado, Mamoru; Kobayashi, Tetsunori; Shirai, Katsuhiko.

Bulletin of Centre for Informatics, Waseda University. Vol. 15 1992. p. 17-23.

Research output: Chapter in Book/Report/Conference proceedingChapter

Endou, T, Watado, M, Kobayashi, T & Shirai, K 1992, Estimation of syntactic structure using prosodic parameters and its application to speech recognition. in Bulletin of Centre for Informatics, Waseda University. vol. 15, pp. 17-23.
Endou T, Watado M, Kobayashi T, Shirai K. Estimation of syntactic structure using prosodic parameters and its application to speech recognition. In Bulletin of Centre for Informatics, Waseda University. Vol. 15. 1992. p. 17-23
Endou, Takashi ; Watado, Mamoru ; Kobayashi, Tetsunori ; Shirai, Katsuhiko. / Estimation of syntactic structure using prosodic parameters and its application to speech recognition. Bulletin of Centre for Informatics, Waseda University. Vol. 15 1992. pp. 17-23
@inbook{c0fe40715bb943c5a52a9e3d65f58b92,
title = "Estimation of syntactic structure using prosodic parameters and its application to speech recognition",
abstract = "Prosodic information is very useful for the estimation of syntactic structure. So Phrase-level segmentation of continuous speech using pitch and power information is discussed to utilize for the word spotting. First, a rule based approach using pitch and power pattern is carried out. Then, a statistical approach is carried out, where the vector sequence of pitch and power are matrix quantized and conditional probabilities of the codes and the labels are calculated. The experimental result of the rule-based approach shows that the boundary detection accuracy is 72.8{\%} in case that the adjacent phrases have loose relation, and 43.4{\%} in case of tight relation. As for the statistical method, the boundary detection accuracy is 63.0{\%}.",
author = "Takashi Endou and Mamoru Watado and Tetsunori Kobayashi and Katsuhiko Shirai",
year = "1992",
month = "6",
language = "English",
volume = "15",
pages = "17--23",
booktitle = "Bulletin of Centre for Informatics, Waseda University",

}

TY - CHAP

T1 - Estimation of syntactic structure using prosodic parameters and its application to speech recognition

AU - Endou, Takashi

AU - Watado, Mamoru

AU - Kobayashi, Tetsunori

AU - Shirai, Katsuhiko

PY - 1992/6

Y1 - 1992/6

N2 - Prosodic information is very useful for the estimation of syntactic structure. So Phrase-level segmentation of continuous speech using pitch and power information is discussed to utilize for the word spotting. First, a rule based approach using pitch and power pattern is carried out. Then, a statistical approach is carried out, where the vector sequence of pitch and power are matrix quantized and conditional probabilities of the codes and the labels are calculated. The experimental result of the rule-based approach shows that the boundary detection accuracy is 72.8% in case that the adjacent phrases have loose relation, and 43.4% in case of tight relation. As for the statistical method, the boundary detection accuracy is 63.0%.

AB - Prosodic information is very useful for the estimation of syntactic structure. So Phrase-level segmentation of continuous speech using pitch and power information is discussed to utilize for the word spotting. First, a rule based approach using pitch and power pattern is carried out. Then, a statistical approach is carried out, where the vector sequence of pitch and power are matrix quantized and conditional probabilities of the codes and the labels are calculated. The experimental result of the rule-based approach shows that the boundary detection accuracy is 72.8% in case that the adjacent phrases have loose relation, and 43.4% in case of tight relation. As for the statistical method, the boundary detection accuracy is 63.0%.

UR - http://www.scopus.com/inward/record.url?scp=0026877143&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0026877143&partnerID=8YFLogxK

M3 - Chapter

AN - SCOPUS:0026877143

VL - 15

SP - 17

EP - 23

BT - Bulletin of Centre for Informatics, Waseda University

ER -