引入包分析的E-Model语音质量评价模型改进

发布时间：2018-07-08 19:20

本文选题：基于IP的语音传输(Vo + IP)　；参考：《重庆邮电大学学报(自然科学版)》2017年02期

【摘要】：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,Vo IP)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Vo IP通信的语音质量。
[Abstract]:In order to evaluate more accurately the damage of voice packet loss to the voice quality of (voice over internet protocol Vo IP based on IP, This paper improves the method of calculating the damage Ie-eff caused by packet loss and coding in the speech quality prediction model E-Model proposed by ITU-T G.107 recommendation. After considering the internal characteristics of speech packet and the existence of burst continuous packet loss, In this paper, the actual loss time (Tloss) is used to measure the speech damage caused by the loss of the speech packet under the fixed speech length. The simulation results show that compared with the original model, the speech quality score obtained by the improved model is compared with the subjective speech quality assessment method (perceptual evaluation of speech quality and PESQ). The Pearson correlation coefficient is increased by 0.045 on average, and the root mean square error is reduced by 0.053 4. The improved E-Model model is more consistent with PESQ in evaluating speech quality, and can more accurately predict the speech quality of Vo IP communication.
【作者单位】：东北电力大学信息工程学院;国网内蒙古东部电力有限公司信息通信分公司;
【分类号】：TN912.3

【相似文献】