Haar-like filtering with center-clipped emphasis for speech detection in sensornet

Jun Nishimura, Tadahiro Kuroda

Research output: Chapter in Book/Report/Conference proceedingConference contribution

3 Citations (Scopus)

Abstract

The use of haar-like filtering for resourced-constrained speech detection in sensornet application is explored. The simple haar-like filters having variable filter width and shift width are trained to learn appropriate filter parameters from the training samples to detect speech. To further refine the accuracy, the center-clipped emphasis is proposed as a new degree of freedom for more adaptive Haar-like filter designs. Our method yielded speech/nonspeech classification accuracy of 98.33% for the input length of 0.1s. Compared with high performance feature extraction method MFCC (Mel-Frequency Cepstrum Coefficient), the proposed haar-like filtering can be approximately 98.40% efficient in terms of the amount of add and multiply computation while capable of achieving the error rate of only 1.63% relative to MFCC.

Original languageEnglish
Title of host publication2009 IEEE 13th Digital Signal Processing Workshop and 5th IEEE Signal Processing Education Workshop, DSP/SPE 2009, Proceedings
Pages1-4
Number of pages4
DOIs
Publication statusPublished - 2009
Event2009 IEEE 13th Digital Signal Processing Workshop and 5th IEEE Signal Processing Education Workshop, DSP/SPE 2009 - Marco Island, FL, United States
Duration: 2009 Jan 42009 Jan 7

Other

Other2009 IEEE 13th Digital Signal Processing Workshop and 5th IEEE Signal Processing Education Workshop, DSP/SPE 2009
CountryUnited States
CityMarco Island, FL
Period09/1/409/1/7

Fingerprint

Feature extraction

Keywords

  • Center-clipped emphasis
  • Haar-like filtering
  • Sensornet
  • Speech detection

ASJC Scopus subject areas

  • Computer Networks and Communications
  • Signal Processing
  • Electrical and Electronic Engineering

Cite this

Nishimura, J., & Kuroda, T. (2009). Haar-like filtering with center-clipped emphasis for speech detection in sensornet. In 2009 IEEE 13th Digital Signal Processing Workshop and 5th IEEE Signal Processing Education Workshop, DSP/SPE 2009, Proceedings (pp. 1-4). [4785885] https://doi.org/10.1109/DSP.2009.4785885

Haar-like filtering with center-clipped emphasis for speech detection in sensornet. / Nishimura, Jun; Kuroda, Tadahiro.

2009 IEEE 13th Digital Signal Processing Workshop and 5th IEEE Signal Processing Education Workshop, DSP/SPE 2009, Proceedings. 2009. p. 1-4 4785885.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Nishimura, J & Kuroda, T 2009, Haar-like filtering with center-clipped emphasis for speech detection in sensornet. in 2009 IEEE 13th Digital Signal Processing Workshop and 5th IEEE Signal Processing Education Workshop, DSP/SPE 2009, Proceedings., 4785885, pp. 1-4, 2009 IEEE 13th Digital Signal Processing Workshop and 5th IEEE Signal Processing Education Workshop, DSP/SPE 2009, Marco Island, FL, United States, 09/1/4. https://doi.org/10.1109/DSP.2009.4785885
Nishimura J, Kuroda T. Haar-like filtering with center-clipped emphasis for speech detection in sensornet. In 2009 IEEE 13th Digital Signal Processing Workshop and 5th IEEE Signal Processing Education Workshop, DSP/SPE 2009, Proceedings. 2009. p. 1-4. 4785885 https://doi.org/10.1109/DSP.2009.4785885
Nishimura, Jun ; Kuroda, Tadahiro. / Haar-like filtering with center-clipped emphasis for speech detection in sensornet. 2009 IEEE 13th Digital Signal Processing Workshop and 5th IEEE Signal Processing Education Workshop, DSP/SPE 2009, Proceedings. 2009. pp. 1-4
@inproceedings{9fecc6539eec477c860a1f2d7787e832,
title = "Haar-like filtering with center-clipped emphasis for speech detection in sensornet",
abstract = "The use of haar-like filtering for resourced-constrained speech detection in sensornet application is explored. The simple haar-like filters having variable filter width and shift width are trained to learn appropriate filter parameters from the training samples to detect speech. To further refine the accuracy, the center-clipped emphasis is proposed as a new degree of freedom for more adaptive Haar-like filter designs. Our method yielded speech/nonspeech classification accuracy of 98.33{\%} for the input length of 0.1s. Compared with high performance feature extraction method MFCC (Mel-Frequency Cepstrum Coefficient), the proposed haar-like filtering can be approximately 98.40{\%} efficient in terms of the amount of add and multiply computation while capable of achieving the error rate of only 1.63{\%} relative to MFCC.",
keywords = "Center-clipped emphasis, Haar-like filtering, Sensornet, Speech detection",
author = "Jun Nishimura and Tadahiro Kuroda",
year = "2009",
doi = "10.1109/DSP.2009.4785885",
language = "English",
isbn = "9781424436774",
pages = "1--4",
booktitle = "2009 IEEE 13th Digital Signal Processing Workshop and 5th IEEE Signal Processing Education Workshop, DSP/SPE 2009, Proceedings",

}

TY - GEN

T1 - Haar-like filtering with center-clipped emphasis for speech detection in sensornet

AU - Nishimura, Jun

AU - Kuroda, Tadahiro

PY - 2009

Y1 - 2009

N2 - The use of haar-like filtering for resourced-constrained speech detection in sensornet application is explored. The simple haar-like filters having variable filter width and shift width are trained to learn appropriate filter parameters from the training samples to detect speech. To further refine the accuracy, the center-clipped emphasis is proposed as a new degree of freedom for more adaptive Haar-like filter designs. Our method yielded speech/nonspeech classification accuracy of 98.33% for the input length of 0.1s. Compared with high performance feature extraction method MFCC (Mel-Frequency Cepstrum Coefficient), the proposed haar-like filtering can be approximately 98.40% efficient in terms of the amount of add and multiply computation while capable of achieving the error rate of only 1.63% relative to MFCC.

AB - The use of haar-like filtering for resourced-constrained speech detection in sensornet application is explored. The simple haar-like filters having variable filter width and shift width are trained to learn appropriate filter parameters from the training samples to detect speech. To further refine the accuracy, the center-clipped emphasis is proposed as a new degree of freedom for more adaptive Haar-like filter designs. Our method yielded speech/nonspeech classification accuracy of 98.33% for the input length of 0.1s. Compared with high performance feature extraction method MFCC (Mel-Frequency Cepstrum Coefficient), the proposed haar-like filtering can be approximately 98.40% efficient in terms of the amount of add and multiply computation while capable of achieving the error rate of only 1.63% relative to MFCC.

KW - Center-clipped emphasis

KW - Haar-like filtering

KW - Sensornet

KW - Speech detection

UR - http://www.scopus.com/inward/record.url?scp=63649090742&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=63649090742&partnerID=8YFLogxK

U2 - 10.1109/DSP.2009.4785885

DO - 10.1109/DSP.2009.4785885

M3 - Conference contribution

SN - 9781424436774

SP - 1

EP - 4

BT - 2009 IEEE 13th Digital Signal Processing Workshop and 5th IEEE Signal Processing Education Workshop, DSP/SPE 2009, Proceedings

ER -