Efficient Frame Erasure Concealment in Predictive Speech Codecs using Glottal Pulse Resynchronisation

Error propagation after a frame loss is an important factor in quality degradation for predictive speech coders. This is mainly due to the lack of synchronization in the adaptive codebook in the good frames following a frame erasure. This article presents a method for resynchronizing the glottal pulse after an erased frame. The method uses an extra frame delay at the decoder, and can be applied with or without additional side information. The approach has been implemented in the G.729.1 standard and results in improved decoder convergence after erased frames. Subjective tests have demonstrated that this improves perceived quality in the presence of frame erasures.