A Self-tuned Quality-Aware De-jittering Buffer Scheme of Packetized Voice Conversations

Transport IP-based networks convey data packets toward their destination with variable one-way network delays due to their best-effort feature. Network delay variation of media packets, known also as network delay jitter, crucially disturbs the perceptual quality of packetized delay-sensitive services such as VoIP, IPTV, and video conferencing. The media-playing entity should seamlessly hide the perceptual effects due to network delay jitter. Such a mechanism is often referred to as de-jittering scheme, which should optimize the lateness-loss / total delay trade-off. In this paper, we describe a novel de-jittering algorithm for VoIP packet streams. The algorithm accommodates a wide range of network delay jitters. This is achieved with the help of three self-tuned first-order filters of experienced one-way network delays and delay variance. Following a new end-to-end delay adjustment event, the de-jittering algorithm tunes for the safety factor, which is used to calibrate the variance around the estimated average network delay, in order to optimize the perceptual quality for a given network condition. The calibration process is performed using the recent history of estimated one-way network delays. This enables to predict with high accuracy the optimal end-to-end delay which will be used till the next adjustment event. Simulation results of VoIPoW (VoIP over wireless) show that our de-jittering algorithm significantly outperforms the baseline de-jittering algorithm [5] in terms of instantaneous and overall perceptual quality.