On the Design and Implementation of Maximum SNR Beamformers for Scattered Speech Sources

Fan Zhang, Chao Pan, Jacob Benesty, Jingdong Chen

科研成果: 书/报告/会议事项章节会议稿件同行评审

摘要

Most existing microphone array beamforming methods for speech signal extraction assume that speech sources obey the point source model. However, this assumption is not applicable in many practical scenarios where small-spacing and small-size microphone arrays are employed, thereby rendering the size of the sources non-negligible. To deal with this situation, the scattered source model should be used, which is investigated in this work. With this model, we derive two versions of the maximum signal-to-noise ratio (SNR) beamformer: one maximizes the SNR with minimum mean-squared error (MSE) and the other maximizes the SNR with minimum speech distortion. We show that the one with minimum MSE can be decomposed as the product between the one with minimum speech distortion and a single-channel Wiener postfilter. We then propose a multichannel approach to estimating the postfilter through noise field modeling. The performance evaluation of the developed beamformers and postfilter is conducted via simulations.

源语言英语
主期刊名Proceedings of 2023 IEEE International Conference on Signal Processing, Communications and Computing, ICSPCC 2023
出版商Institute of Electrical and Electronics Engineers Inc.
ISBN(电子版)9798350316728
DOI
出版状态已出版 - 2023
活动2023 IEEE International Conference on Signal Processing, Communications and Computing, ICSPCC 2023 - Zhengzhou, Henan, 中国
期限: 14 11月 202317 11月 2023

出版系列

姓名Proceedings of 2023 IEEE International Conference on Signal Processing, Communications and Computing, ICSPCC 2023

会议

会议2023 IEEE International Conference on Signal Processing, Communications and Computing, ICSPCC 2023
国家/地区中国
Zhengzhou, Henan
时期14/11/2317/11/23

指纹

探究 'On the Design and Implementation of Maximum SNR Beamformers for Scattered Speech Sources' 的科研主题。它们共同构成独一无二的指纹。

引用此