An objective measurement of speech quality in VoIP based on network parameters

With the development of IP technology, the popularity of VoIP voice service is increasing fast. In the fourth generation mobile communication system, the voice service is transformed into IP service, and transmits voice data using IP network. However, due to the feature of the best effort service in packet-based network, many network parameters such as delay, jitter, packet loss and so on, will have a great impact on the quality of VoIP voice. In order to assess the voice quality of VoIP accurately, and to avoid the problem that the subjective test methods consume a large amount of human resources, this paper presents a novel non-intrusive objective evaluation method of VoIP speech quality based on the network parameters. Based on the modeling methods of multiple regression and neural network with principal component analysis, the training and verification of the model are evaluated based on the data captured from IP network. The experimental results show that, in case of sufficient training data, compared with the PESQ method, the objective quality scores obtained by the proposed evaluation method have a high correlation with the scores obtained by POLQA which is currently the optimal performance method in VoIP speech quality evaluation.