Algoritmo para melhorar o desempenho de uma métrica não intrusiva de qualidade de voz

Abstract

Telephone systems are an important communication tool nowadays, and as such, have their quality constantly questioned. In order to provide evaluation mechanisms, the Telecommunication Standardization Sector of the International Telecommunication Union (ITU-T) establishes technical recommendations for determining the measure metrics of the quality of experience (QoE).The ITU-T recommendations P.862 and P.563 stand out among the others, and through the Mean Opinion Score (MOS) they assign a value for the audio sample being evaluated. Thus, the work introduces a solution for assessing voice quality in real-time telecommunication services, which is useful for subscribers, mobile service providers (MSP), and telecommunication regulatory authorities (TRA). The main objective of this research is to provide a model that improves the performance of the algorithm available in the ITU-T Rec. P.563. To accomplish this objective, the proposed model considers two factors: a better response in network conditions with packet loss and a more adequate treatment of silences segments present into the audio signal. Thus, the proposed model gives rise to a Voice Quality Server (VQS) that is implemented in a Web Service in order to provide conditions for the TRAs and MSPs to monitor the state of the network, specifically the values of the MOS index of the established Voice Over IP (VoIP) calls. In addition, this solution is implemented in mobile application format and is called VQSmobile . Such an approach then allows for the making of calls and the proper presentation of the MOS index value to the user. The experimental results show that the P.563 algorithm performance was really improved by the proposed model, approximating its results to those given by ITU-T Rec. P.862 algorithm, reaching a Pearson correlation coefficient (PCC) of 0.9957 and a root-mean square error (RMSE) of 0.2983. Moreover, the processing and energy consumption corresponding to the application are almost negligible in current mobile devices. In addition, the proposed solution provides an important tool for users to monitor the quality of the services contracted with the MSP.