面向单传声器节点的分布式语音增强算法  

Distributed speech enhancement with single microphone per node

作　　者：庞淑蓉 秦品乐[1] 曾建潮 畅瑞江 PANG Shurong;QIN Pinle;ZENG Jianchao;CHANG Ruijiang(School of Computer Science and Technology,North University of China,Taiyuan 030051,China;School of Electrical Engineering,Hebei University of Architecture,Zhangjiakou 075000,Hebei China)

机构地区：[1]中北大学计算机科学与技术学院,山西太原030051 [2]河北建筑工程学院电气工程学院,河北张家口075000

出　　处：《华中科技大学学报(自然科学版)》2025年第3期99-108,共10页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：山西科技厅重大专项基金资助项目(202101010101018);山西省基础研究计划青年项目(202403021212105);河北建筑工程学院博士启动基金资助项目(B-202406)。

摘　　要：在节点为单传声器的无线声学传感器网络(WASN)中,针对分布式语音增强时声学传递函数难以获取及节点间频繁交换数据的问题,提出了一种基于单传声器节点的分布式语音增强算法.具体地,首先针对全连接WASN,提出了基于最大输入信噪比节点的语音失真加权分布式维纳滤波(SDW-DWF)算法.该算法通过计算获得最大输入信噪比节点,并由该节点接收其他节点发送的多帧信号信息,来分布式实现加权维纳滤波语音增强.该算法在节点间仅需交换少量数据,且不依赖于声学传递函数.此外,提出了一种无拓扑约束的比较一致性算法,来获得最大输入信噪比节点,以使得SDW-DWF方法可用于无网络拓扑约束的移动说话人场景.仿真和实际实验验证了本文方法在无需先验约束情况下,具有较高的分布式语音增强性能.To address the issues of difficult acoustic transfer function acquisition and frequent data exchange among nodes for distributed speech enhancement,a distributed speech enhancement method was proposed for the wireless acoustic sensor network(WASN) with a single microphone per node.Specifically,based on the node with the maximum input signal-to-noise ratio,the speech distortion weighted distributed Wiener filter(SDW-DWF) was first proposed in the fully connected WASN.To achieve distributed weighted Wiener filtering speech enhancement,the maximum input signal-to-noise ratio node needs to be obtained,and the node receives multiple frames of signal information sent by other nodes.The proposed method only requires data exchange among nodes once and does not rely on the acoustic transfer function.In addition,to obtain the node with the maximum input signalto-noise ratio in any topology,the comparative consensus algorithm was proposed,so that the SDW-DWF method can be used in mobile speaker scenarios without topology constraint.Simulation and real-world experimental results illustrate that the proposed method without any prior information achieves good speech enhancement performance.

关 键 词：分布式语音增强 分布式维纳滤波 加权维纳滤波 一致性算法 无线声学传感器网络 

分 类 号：TN912.35[电子电信—通信与信息系统]