Soft Counting Poisson Mixture Model-Based Polling Method for Speech/Nonspeech Classification

Youngjoo SUH; Hoirin KIM; Minsoo HAHN; Yongju LEE

doi:10.1093/ietisy/e89-d.12.2994

Soft Counting Poisson Mixture Model-Based Polling Method for Speech/Nonspeech Classification

Youngjoo SUH, Hoirin KIM, Minsoo HAHN, Yongju LEE

Full Text Views

0

Cite this

Summary :

In this letter, a new segment-level speech/nonspeech classification method based on the Poisson polling technique is proposed. The proposed method makes two modifications from the baseline Poisson polling method to further improve the classification accuracy. One of them is to employ Poisson mixture models to more accurately represent various segmental patterns of the observed frequencies for frame-level input features. The other is the soft counting-based frequency estimation to improve the reliability of the observed frequencies. The effectiveness of the proposed method is confirmed by the experimental results showing the maximum error reduction of 39% compared to the segmentally accumulated log-likelihood ratio-based method.

Publication: IEICE TRANSACTIONS on Information Vol.E89-D No.12 pp.2994-2997

Publication Date: 2006/12/01

Publicized

Online ISSN: 1745-1361

DOI: 10.1093/ietisy/e89-d.12.2994

Type of Manuscript: LETTER

Category: Speech and Hearing

Cite this

Copy

Youngjoo SUH, Hoirin KIM, Minsoo HAHN, Yongju LEE, "Soft Counting Poisson Mixture Model-Based Polling Method for Speech/Nonspeech Classification" in IEICE TRANSACTIONS on Information, vol. E89-D, no. 12, pp. 2994-2997, December 2006, doi: 10.1093/ietisy/e89-d.12.2994.
Abstract: In this letter, a new segment-level speech/nonspeech classification method based on the Poisson polling technique is proposed. The proposed method makes two modifications from the baseline Poisson polling method to further improve the classification accuracy. One of them is to employ Poisson mixture models to more accurately represent various segmental patterns of the observed frequencies for frame-level input features. The other is the soft counting-based frequency estimation to improve the reliability of the observed frequencies. The effectiveness of the proposed method is confirmed by the experimental results showing the maximum error reduction of 39% compared to the segmentally accumulated log-likelihood ratio-based method.
URL: https://global.ieice.org/en_transactions/information/10.1093/ietisy/e89-d.12.2994/_p

Copy

@ARTICLE{e89-d_12_2994,
author={Youngjoo SUH, Hoirin KIM, Minsoo HAHN, Yongju LEE, },
journal={IEICE TRANSACTIONS on Information},
title={Soft Counting Poisson Mixture Model-Based Polling Method for Speech/Nonspeech Classification},
year={2006},
volume={E89-D},
number={12},
pages={2994-2997},
abstract={In this letter, a new segment-level speech/nonspeech classification method based on the Poisson polling technique is proposed. The proposed method makes two modifications from the baseline Poisson polling method to further improve the classification accuracy. One of them is to employ Poisson mixture models to more accurately represent various segmental patterns of the observed frequencies for frame-level input features. The other is the soft counting-based frequency estimation to improve the reliability of the observed frequencies. The effectiveness of the proposed method is confirmed by the experimental results showing the maximum error reduction of 39% compared to the segmentally accumulated log-likelihood ratio-based method.},
keywords={},
doi={10.1093/ietisy/e89-d.12.2994},
ISSN={1745-1361},
month={December},}

Copy

TY - JOUR
TI - Soft Counting Poisson Mixture Model-Based Polling Method for Speech/Nonspeech Classification
T2 - IEICE TRANSACTIONS on Information
SP - 2994
EP - 2997
AU - Youngjoo SUH
AU - Hoirin KIM
AU - Minsoo HAHN
AU - Yongju LEE
PY - 2006
DO - 10.1093/ietisy/e89-d.12.2994
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E89-D
IS - 12
JA - IEICE TRANSACTIONS on Information
Y1 - December 2006
AB - In this letter, a new segment-level speech/nonspeech classification method based on the Poisson polling technique is proposed. The proposed method makes two modifications from the baseline Poisson polling method to further improve the classification accuracy. One of them is to employ Poisson mixture models to more accurately represent various segmental patterns of the observed frequencies for frame-level input features. The other is the soft counting-based frequency estimation to improve the reliability of the observed frequencies. The effectiveness of the proposed method is confirmed by the experimental results showing the maximum error reduction of 39% compared to the segmentally accumulated log-likelihood ratio-based method.
ER -