一种基于图后置滤波的多通道语音增强方法研究
A multi-channel speech enhancement method based on graph post-filtering
  
DOI:
中文关键词:  图信号处理;多通道语音增强;图拓扑结构;后置滤波
英文关键词:graph signal processing (GSP); multi channel speech enhancement; graph topological structure; post filtering
基金项目:国家自然科学基金(62071242)资助项目
作者单位
张鹏程 南京邮电大学 通信与信息工程学院,江苏 南京 210003 
郭海燕 南京邮电大学 通信与信息工程学院,江苏 南京 210004 
杨 震 南京邮电大学 通信与信息工程学院,江苏 南京 210005
南京邮电大学 通信与网络技术国家地方联合工程研究中心,江苏 南京 210003 
杨 洋 南京邮电大学 通信与信息工程学院,江苏 南京 210006 
摘要点击次数: 260
全文下载次数: 92
中文摘要:
      结合图信号处理与波束形成,提出了一种新的图后置滤波(Graph Post Filtering, GPF)方法来实现多通道语音信号增强。首先,提出了多阶自旋的图拓扑结构来构建语音图信号。然后,针对复杂散射噪声场下维纳后置滤波(Wiener Post Filtering, WPF)方法的局限性,结合波束形成提出了一种新的图后置滤波方法,基于输入各路语音图信号的自相关和互相关功率谱密度,推导出增益函数,进而预测源语音信号的图频谱。实验结果表明,所提出的GPF方法在信噪比(Signal to Noise Ratio, SNR)和主观语音质量评估(Perceptual Evaluation of Speech Quality, PESQ)方面,性能均优于传统的WPF方法。另外,实验结果也表明,各路信号时延补偿偏差会影响基于GPF的多通道语音增强性能。
英文摘要:
      In this paper, we propose a novel graph post filtering (GPF) method to enhance multi channel speech signals by combining graph signal processing and beamforming. First, we design a multi order self spin graph topology to construct speech graph signals. Second, considering the limitations of Wiener post filtering (WPF) method in complex scattered noise fields, we propose a novel GPF method combined with beamforming. Specifically, based on the autocorrelation and cross correlation power spectrum density of the input speech signal, the gain function is derived to predict the graph spectrum of the source speech signal. Our experimental results show that the proposed GPF method outperforms the traditional WPF method in terms of both signal to noise ratio (SNR) and perceptual evaluation of speech quality (PESQ). The results also demonstrate that the delay compensation deviation of each channel will affect the performance of GPF based multi channel speech enhancement.
查看全文  查看/发表评论  下载PDF阅读器

你是第2812143访问者
版权所有《南京邮电大学学报(自然科学版)》编辑部
Tel:86-25-85866913 E-mail:xb@njupt.edu.cn
技术支持:本系统由北京勤云科技发展有限公司设计