Intra-flow loss recovery and control for VoIP
MULTIMEDIA '01 Proceedings of the ninth ACM international conference on Multimedia
NOSSDAV '02 Proceedings of the 12th international workshop on Network and operating systems support for digital audio and video
Hi-index | 0.00 |
Losing VoIP packets or speech frames decreases the perceptual speech quality. The statistical relation between randomly lost speech frames and speech quality is well known. In cases of bursty and rate-distortion optimized losses, a precise quality model is required to relate losses to quality. In the present paper, we present a model that is based on the loss impact – or the importance – of single speech frames. We present a novel metric to calculate the impact of the loss of multiple frames by adding the importance of the respective single frames. This metric shows a high prediction accuracy for distant losses. For losses following each other closely, we present an aggregation function which models the psychoacoustic post-masking effect. Our model helps to develop networking algorithms that control the packet dropping process in networks. For example, we show that a proper packet dropping strategy can significantly increase the drop rate while maintaining the same level of speech quality.