引入包分析的E-Model语音质量评价模型改进

为了能够更加准确地评价语音包丢失对基于IP的语音传输(voice over internet protocol,VoIP)的语音质量的损伤,对ITU-T G.107建议书提出的语音质量预测模型E-Model中计算丢包与编码造成的损伤Ie-eff的方法作出改进,在综合考虑语音包的内部特性和存在突发连续丢包情况后,提出利用在固定语音长度下,语音实际损失时间Tloss来衡量语音包丢失造成的语音损伤.仿真结果表明,相比原有模型,改进后的模型得到的语音质量评分同主观语音质量评估方法(perceptual evaluation of speech quality,PESQ)评分相比,皮尔森相关系数平均提高了0.045 8,均方根误差平均降低了0.053 4,改进后的E-Model模型在评价语音质量时与PESQ更具有一致性,可以更为准确地预测VoIP通信的语音质量.

Improved E-Model of speech quality evaluation model based on packet analysis

In order to evaluate the voice quality damage of IP-based voice over internet protocol (VoIP) caused by the loss of voice packets more accurately,we improved the method of calculating Ie-eff,(the damage caused by encoding,decoding and packet loss) in E-Model proposed by ITU-T G.107 recommendations.After considering the internal characteristics of voice package and the situation of unexpected continuous packet loss,we proposed to use Tloss,the actual loss time of voice using fixed voice length,to measure the voice damage caused by the loss of voice packet.The simulation results show that compared with the original model,the pearson's correlation coefficient between the voice quality value using the improved model and the value using perceptual evaluation of speech quality (PESQ) are improved by an average of 0.045 8,and the RMSE are reduced by an average of 0.053 4.Therefore,the improved E-Model is more consistent with PESQ,and more accurate to predict the quality of VoIP voice communication.