Towards degradation decomposition for voice communication system assessment
Towards degradation decomposition for voice communication system assessment
复制标题
语音通信系统评估的退化分解
DOI：
--
复制
发表时间：
2017
期刊：
影响因子：
--
通讯作者：
L. Malfait
中科院分区：
文献类型：
--
作者：
Friedemann Köster;F. Schiffner;Sebastian Möller;L. Malfait
本文介绍了用于评估语音通信的退化分解工具的最新发展。总体质量分数，即由ITU-T P.800绝对类别评级（ACR）等主观测试方法产生的平均意见分数（MOS），仍然是业界最流行的质量指标。虽然MOS是评估质量问题的一个很好的指标，但它不能提供有关问题原因的信息。为了解决这一差距，ITU-T目前正在开展工作项目，为业界提供通过感知或技术退化分解来了解较低分数的原因的方法。目标是生成支持自动降解分解的客观模型。这种开发的第一步是构建用于模型训练和验证的数据库。为此，总共使用ITU-T讨论的电位诊断测试方法进行了四个实验。此外，提出并讨论了测试方法的两个可选改进。实验结果表明，分析测试方法在标准化方面仍有验证和进一步改进的空间。
This article presents the current development of degradation decomposition tools for the assessment of voice communications. Overall quality scores, represented as Mean Opinion Scores (MOS) produced by subjective test methodologies such as ITU-T P.800 Absolute Category Rating (ACR), remains the most popular quality metric in the industry. While MOS is a great indicator to evaluate quality issues, it does not provide information on the cause of issues. To address this gap, work items are currently active within ITU-T to provide the industry with means to understand the cause of lower scores by perceptual or technical degradation decompositions. The goal is to produce objective models that enable automated degradation decomposition. The first step in such a development is the construction of databases for model training and validation. For this, in sum four experiments using a potential diagnostic test method discussed within ITU-T are conducted. In addition, two optional improvements for the test method are presented and discussed. The results of the experiments show that for standardization the analyzed test method still leaves room for validation and further improvements.