Hybrid real-time quality assessment model for voice over IP

Abstract

Call quality assessment is of significant importance for adaptive multimedia transmission in order to support high quality of delivery of multimedia services in current dynamic network environments. Due to the high costs in terms of time and human resources of performing subjective call quality tests, objective quality assessment methods are preferred. Intrusive objective methods are reference-based and speech content dependent, but cannot be done in real-time; non-intrusive objective methods are suitable for real-time quality monitoring, but their results do not reflect the speech content, which is highly related to user quality of experience. In this paper, a novel Hybrid Call Quality (HCQ) assessment model is proposed which combines the benefits of the existing solutions. The HCQ model estimates the call quality using a non-intrusive method at the beginning of the call or voice transmission, and then corrects itself using an intrusive method. The proposed model uses weights when combining the results of the two methods, which are derived from an extensive set of subjective results. Results show that when employing the HCQ model there is higher correlation with the subjective results in comparison with using existing methods.