No-reference quality assessment model for networked speech based on the content feature of packets
Fuzheng Yang · Journal of Xidian University · 2011
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.