文章摘要
朱升伟,王杰,厉剑,李晓东,郑成诗.分布式加权功率最小化无失真响应联合去噪去混响波束形成器*[J].,2025,44(4):955-970
分布式加权功率最小化无失真响应联合去噪去混响波束形成器*
Distributed weighted power minimization distortionless response joint denoising and dereverberation beamformer
投稿时间:2024-02-06  修订日期:2025-06-29
中文摘要:
      在室内场景中,传声器拾取的语声会受到噪声和混响的严重影响。为了改善语声质量与可懂度,提出了一种适用于无线声传感网络的基于加权功率最小化无失真响应的分布式联合去噪去混响算法。该文首先推导了分布式阵列场景下基于加权功率最小化无失真响应约束的联合去噪去混响波束形成器的解析表达式,并且对该方法的收敛性进行分析。在所提方法中,各个节点首先将本地节点的多通道信号压缩为单通道信号,并将该压缩信号发送至其他所有节点。随后,根据加权功率最小化无失真准则,每个节点利用本地节点信号和其他节点的压缩信号计算滤波器权重,显著降低了节点间的传输带宽和节点内的计算复杂度。仿真实验结果表明,所提方法在各项评价指标中均优于其他分布式算法,可以在计算复杂度和传输带宽显著降低的条件下实现与集中式算法相当的去噪去混响效果。
英文摘要:
      In indoor environments, the quality of speech captured by a microphone can be significantly affected by noise and reverberation. In order to improve speech quality and intelligibility, this paper proposes a distributed joint denoising and dereverberation algorithm based on weighted power minimization distortionless response for wireless acoustic sensor networks (WASNs). We begin by deriving the analytical expression of the joint denoising and dereverberation beamformer under the weighted power minimization distortionless response constraint in a distributed array scenario, and analyze its convergence properties. In the proposed method, each node compresses the multi-channel signal of the local node into a single-channel signal, which is then shared with all other nodes. Employing the local node signals and the compressed signals from other nodes, each node calculates the weight vectors of the filter according to the weighted power minimization distortionless criterion. Simulation results demonstrate the superiority of the proposed method over other distributed algorithms across various evaluation metrics, while achieving comparable performance to centralized algorithms in terms of denoising and dereverberation. Moreover, the proposed method significantly reduces both the transmission bandwidth required between nodes and the computational complexity within each node.
DOI:10.11684/j.issn.1000-310X.2025.04.017
中文关键词: 分布式阵列  去噪去混响  波束形成
英文关键词: Distributed array  Denoising and dereverberation  Beamformer
基金项目:国家自然科学基金项目(面上项目,重点项目,重大项目)
作者单位E-mail
朱升伟 广州大学 2112130090@e.gzhu.edu.cn 
王杰 广州大学 wangjie@gzhu.edu.cn 
厉剑* 中国科学院声学研究所 lijian@mail.ioa.ac.cn 
李晓东 中国科学院声学研究所 lxd@mail.ioa.ac.cn 
郑成诗 中国科学院声学研究所 cszheng@mail.ioa.ac.cn 
摘要点击次数: 607
全文下载次数: 73
查看全文   查看/发表评论  下载PDF阅读器
关闭