New methods for lost voice packet recovery in active networks

Abstract

This paper presents two new voice-packing schemes based on speech properties for voice over Internet protocol. Those methods aim to reduce the speech quality degradation caused by packet losses. They provide feasibility of regenerating lost voice packets within the network in active nodes. We argue that in a very lossy network, the receiver cannot cope with all effects of lost packets and the voice has poor quality. The two proposed methods could help regenerate lost packets in an active network to improve the perceptual quality of sound. These schemes are evaluated by simulating a very lossy network with adjustable drop rate. The experiments showed that the regenerated voice has much better quality compared with the other compensating techniques even under more than 35% packet loss rate.