Speech coding over frame relay networks

In this paper, the problem of transmitting compressed voice over frame relay networks is addressed. To overcome variable frame delay and frame loss in the networks, an adaptive build out delay mechanism combining silence detection and time-scale modification is described for use in an embedded 8/16 kb/s CELP-MPE speech coder. Vector Linear Prediction (VLP) is then discussed for the purpose of voice frame re-construction. During frame losses, an adaptive energy control mechanism is introduced to suppress sudden energy explosion in the reconstructed frame. Objective and subjective results at random frame error rates are obtained and it is shown that there is virtually no degradation in perceptual quality when the frame loss rate is as high as 1%.