考虑包内容特性的网络语音质量评价模型  

No-reference quality assessment model for networked speech based on the content feature of packets

作　　者：李维[1] 杨付正[1] 

机构地区：[1]西安电子科技大学综合业务网理论及关键技术国家重点实验室,陕西西安710071

出　　处：《西安电子科技大学学报》2011年第2期23-28,98,共7页Journal of Xidian University

基　　金：国家自然科学基金资助项目(60902081;61072102);中央高校基本科研业务费资助项目(K50510010021);国际科技合作计划资助项目(2010DFB10570)

摘　　要：丢包是引起网络语音质量下降的主要因素,其影响程度不仅与丢失包的个数有关,而且与丢失包的内容特性有关.考虑丢失包的内容特性,提出了一种无参考的网络语音质量评价模型.首先,利用话音激活检测技术和未丢帧的电平来判断丢失包的内容特性,然后统计语音包的丢失率,进而提出一种无参考的网络语音质量评价模型来预测网络失真语音质量.实验结果表明,相比于国际标准G.1070中的语音质量评价模型,无参考的网络语音质量评价模型与主观质量评价的相关性平均提高了8.4%.Packet loss is a primary element to impair the quality of network speech,and the impairment extent is related not only to the number of lost packets,but also to their content feature.This paper proposes a no-reference quality assessment model for networked speech based on analysis of the content feature of packets.This model first estimates the content feature of lost packets by resorting to the voice active detection technology and the level of non-lost frame.Based on the content feature analysis,the proposed model calculates the speech packet loss rate and then predicts the quality of networked speech.Experimental results show that high prediction accuracy can be obtained by the proposed model.Compared with the speech quality evaluation model in G.1070,the average performance gain of the proposed model is raised 8.4% in terms of relativity to subjective results.

关 键 词：语音质量 语音质量评价 包丢失 包内容特性 

分 类 号：TN912[电子电信—通信与信息系统]