Jitter Buffer Compensation in Voice over IP Quality Estimation

Voice over Internet Protocol (VoIP) communications has grown in popularity and has been widely adopted as an alternative to traditional telephone technologies. As packet delivery time can vary, congestion and poor connections can result in delays being introduced. In order to maximise the Quality of Experience (QoE) for VoIP users, jitter buffers are deployed to manage the speech output at the receiver by minimising stalls in signal playout. Automated computer models have been created to compare the signal at the origin and destination in order to predict the perceived quality. Although short playout adjustments introduced by jitter buffers are imperceptible to human listeners, they can skew the results from speech quality prediction models. In this paper, the influence of origin and destination signal time alignment is investigated. A new algorithm for delay estimation that can be used for Jitter Buffer Compensation is proposed and evaluated for use within a speech quality prediction model called VISQOL. Two experiments were conducted. The first was used to design a delay estimation algorithm and to tune its parameters. The second validates the algorithm performance by comparing the quality prediction accuracy of the VISQOL speech quality model with the proposed Jitter Buffer Compensation model to the baseline model results. The results show that the proposed algorithm produces significantly fewer signal mis-alignments and better quality prediction.