摘要
语音通话是电信网络下的主流业务之一,业界主流的语音业务评估方案通过对比收/发端语音音频,计算得到POLQA MOS评估业务质量。对于电信运营商,在不侵犯用户隐私的前提下感知业务质量,进而进行针对性的业务质量保障是重点工作之一。提出了通过数理统计及神经网络算法,学习网络侧数据包交互信息与终端侧语音业务质量的映射关系,形成高精度评估模型;在此基础上,结合无线网用户及小区级多维数据信息,分析输出业务质差的原因及解决方案,形成高效准确保障业务质量的系统。实际应用效果证明,该系统可通过网络侧数据准实时完成语音业务质量评估,并输出质差解决方案,针对网络中的VoLTE语音业务进行质量保障工作。
The voice service is one of the mainstream services under the telecommunications network.The mainstream voice service quality evaluation method in the industry is to compare the voice and audio of the transceiver and calculate the POLQA MOS,which is the standard benchmark for voice quality analysis.For telecom operators,perceive voice service quality efficiently without any infringement of user privacy,and carry out precise service quality assurance,is one of the key tasks.A system using mathematical statistics and neural network to learn the mapping between voice service feature data from network side and voice quality from user side to generate evaluation model with high accuracy and precision was proposed.On this basis,the system made use of data information from multiple fields including wireless network user level and cell level,to analyze reasons and work out solutions for poor service quality,so as to assure service quality efficiently and accurately.The result shows that the scheme has achieved good practical application results.
作者
李光宇
梁燕萍
余立
王昆
陈扬铭
刘思佳
LI Guangyu;LIANG Yanping;YU Li;WANG Kun;CHEN Yangming;LIU Sijia(China Mobile Research Institute,Beijing 100053,China)
出处
《电信科学》
2021年第5期64-71,共8页
Telecommunications Science
