Model-Based Voice Activity Detection in Wireless Acoustic Sensor Networks

Research output: Contribution to book/anthology/report/conference proceedingArticle in proceedingResearchpeer-review

60 Downloads (Pure)

Abstract

One of the major challenges in wireless acoustic sensor networks (WASN) based speech enhancement is robust and accurate voice activity detection (VAD). VAD is widely used in speech enhancement, speech coding, speech recognition, etc. In speech enhancement applications, VAD plays an important role, since noise statistics can be updated during non-speech frames to ensure efficient noise reduction and tolerable speech distortion. Although significant efforts have been made in single channel VAD, few solutions can be found in the multichannel case, especially in WASN. In this paper, we introduce a distributed VAD by using model-based noise power spectral density (PSD) estimation. For each node in the network, the speech PSD and noise PSD are first estimated, then a distributed detection is made by applying the generalized likelihood ratio test (GLRT). The proposed global GLRT based VAD has a quite general form. Indeed, we can judge whether the speech is present or absent by using the current time frame and frequency band observation or by taking into account the neighbouring frames and bands. Finally, the distributed GLRT result is obtained by using a distributed consensus method, such as random gossip, i.e., the whole detection system does not need any fusion center. With the model-based noise estimation method, the proposed distributed VAD performs robustly under non-stationary noise conditions, such as babble noise. As shown in experiments, the proposed method outperforms traditional multichannel VAD methods in terms of detection accuracy.
Original languageEnglish
Title of host publication2018 26th European Signal Processing Conference (EUSIPCO)
Number of pages5
PublisherIEEE
Publication dateSep 2018
Pages425-429
Article number8553457
ISBN (Print)978-90-827970-0-8, 978-1-5386-3736-4
ISBN (Electronic)978-9-0827-9701-5
DOIs
Publication statusPublished - Sep 2018
Event26th European Signal Processing Conference (EUSIPCO 2018) - Rome, Italy
Duration: 3 Sep 20187 Sep 2018
Conference number: 26
http://www.eusipco2018.org

Conference

Conference26th European Signal Processing Conference (EUSIPCO 2018)
Number26
CountryItaly
CityRome
Period03/09/201807/09/2018
Internet address
SeriesProceedings of the European Signal Processing Conference
ISSN2076-1465

Cite this

Zhao, Y., Nielsen, J. K., Christensen, M. G., & Chen, J. (2018). Model-Based Voice Activity Detection in Wireless Acoustic Sensor Networks. In 2018 26th European Signal Processing Conference (EUSIPCO) (pp. 425-429). [8553457] IEEE. Proceedings of the European Signal Processing Conference https://doi.org/10.23919/EUSIPCO.2018.8553457
Zhao, Yingke ; Nielsen, Jesper Kjær ; Christensen, Mads Græsbøll ; Chen, Jingdong. / Model-Based Voice Activity Detection in Wireless Acoustic Sensor Networks. 2018 26th European Signal Processing Conference (EUSIPCO). IEEE, 2018. pp. 425-429 (Proceedings of the European Signal Processing Conference).
@inproceedings{318a8053391d43a1ad842ab39729b72d,
title = "Model-Based Voice Activity Detection in Wireless Acoustic Sensor Networks",
abstract = "One of the major challenges in wireless acoustic sensor networks (WASN) based speech enhancement is robust and accurate voice activity detection (VAD). VAD is widely used in speech enhancement, speech coding, speech recognition, etc. In speech enhancement applications, VAD plays an important role, since noise statistics can be updated during non-speech frames to ensure efficient noise reduction and tolerable speech distortion. Although significant efforts have been made in single channel VAD, few solutions can be found in the multichannel case, especially in WASN. In this paper, we introduce a distributed VAD by using model-based noise power spectral density (PSD) estimation. For each node in the network, the speech PSD and noise PSD are first estimated, then a distributed detection is made by applying the generalized likelihood ratio test (GLRT). The proposed global GLRT based VAD has a quite general form. Indeed, we can judge whether the speech is present or absent by using the current time frame and frequency band observation or by taking into account the neighbouring frames and bands. Finally, the distributed GLRT result is obtained by using a distributed consensus method, such as random gossip, i.e., the whole detection system does not need any fusion center. With the model-based noise estimation method, the proposed distributed VAD performs robustly under non-stationary noise conditions, such as babble noise. As shown in experiments, the proposed method outperforms traditional multichannel VAD methods in terms of detection accuracy.",
author = "Yingke Zhao and Nielsen, {Jesper Kj{\ae}r} and Christensen, {Mads Gr{\ae}sb{\o}ll} and Jingdong Chen",
year = "2018",
month = "9",
doi = "10.23919/EUSIPCO.2018.8553457",
language = "English",
isbn = "978-90-827970-0-8",
series = "Proceedings of the European Signal Processing Conference",
publisher = "IEEE",
pages = "425--429",
booktitle = "2018 26th European Signal Processing Conference (EUSIPCO)",
address = "United States",

}

Zhao, Y, Nielsen, JK, Christensen, MG & Chen, J 2018, Model-Based Voice Activity Detection in Wireless Acoustic Sensor Networks. in 2018 26th European Signal Processing Conference (EUSIPCO)., 8553457, IEEE, Proceedings of the European Signal Processing Conference, pp. 425-429, 26th European Signal Processing Conference (EUSIPCO 2018), Rome, Italy, 03/09/2018. https://doi.org/10.23919/EUSIPCO.2018.8553457

Model-Based Voice Activity Detection in Wireless Acoustic Sensor Networks. / Zhao, Yingke; Nielsen, Jesper Kjær; Christensen, Mads Græsbøll; Chen, Jingdong.

2018 26th European Signal Processing Conference (EUSIPCO). IEEE, 2018. p. 425-429 8553457 (Proceedings of the European Signal Processing Conference).

Research output: Contribution to book/anthology/report/conference proceedingArticle in proceedingResearchpeer-review

TY - GEN

T1 - Model-Based Voice Activity Detection in Wireless Acoustic Sensor Networks

AU - Zhao, Yingke

AU - Nielsen, Jesper Kjær

AU - Christensen, Mads Græsbøll

AU - Chen, Jingdong

PY - 2018/9

Y1 - 2018/9

N2 - One of the major challenges in wireless acoustic sensor networks (WASN) based speech enhancement is robust and accurate voice activity detection (VAD). VAD is widely used in speech enhancement, speech coding, speech recognition, etc. In speech enhancement applications, VAD plays an important role, since noise statistics can be updated during non-speech frames to ensure efficient noise reduction and tolerable speech distortion. Although significant efforts have been made in single channel VAD, few solutions can be found in the multichannel case, especially in WASN. In this paper, we introduce a distributed VAD by using model-based noise power spectral density (PSD) estimation. For each node in the network, the speech PSD and noise PSD are first estimated, then a distributed detection is made by applying the generalized likelihood ratio test (GLRT). The proposed global GLRT based VAD has a quite general form. Indeed, we can judge whether the speech is present or absent by using the current time frame and frequency band observation or by taking into account the neighbouring frames and bands. Finally, the distributed GLRT result is obtained by using a distributed consensus method, such as random gossip, i.e., the whole detection system does not need any fusion center. With the model-based noise estimation method, the proposed distributed VAD performs robustly under non-stationary noise conditions, such as babble noise. As shown in experiments, the proposed method outperforms traditional multichannel VAD methods in terms of detection accuracy.

AB - One of the major challenges in wireless acoustic sensor networks (WASN) based speech enhancement is robust and accurate voice activity detection (VAD). VAD is widely used in speech enhancement, speech coding, speech recognition, etc. In speech enhancement applications, VAD plays an important role, since noise statistics can be updated during non-speech frames to ensure efficient noise reduction and tolerable speech distortion. Although significant efforts have been made in single channel VAD, few solutions can be found in the multichannel case, especially in WASN. In this paper, we introduce a distributed VAD by using model-based noise power spectral density (PSD) estimation. For each node in the network, the speech PSD and noise PSD are first estimated, then a distributed detection is made by applying the generalized likelihood ratio test (GLRT). The proposed global GLRT based VAD has a quite general form. Indeed, we can judge whether the speech is present or absent by using the current time frame and frequency band observation or by taking into account the neighbouring frames and bands. Finally, the distributed GLRT result is obtained by using a distributed consensus method, such as random gossip, i.e., the whole detection system does not need any fusion center. With the model-based noise estimation method, the proposed distributed VAD performs robustly under non-stationary noise conditions, such as babble noise. As shown in experiments, the proposed method outperforms traditional multichannel VAD methods in terms of detection accuracy.

U2 - 10.23919/EUSIPCO.2018.8553457

DO - 10.23919/EUSIPCO.2018.8553457

M3 - Article in proceeding

SN - 978-90-827970-0-8

SN - 978-1-5386-3736-4

T3 - Proceedings of the European Signal Processing Conference

SP - 425

EP - 429

BT - 2018 26th European Signal Processing Conference (EUSIPCO)

PB - IEEE

ER -

Zhao Y, Nielsen JK, Christensen MG, Chen J. Model-Based Voice Activity Detection in Wireless Acoustic Sensor Networks. In 2018 26th European Signal Processing Conference (EUSIPCO). IEEE. 2018. p. 425-429. 8553457. (Proceedings of the European Signal Processing Conference). https://doi.org/10.23919/EUSIPCO.2018.8553457