Channel error resilience algorithm for vocoder LSP parameters
Tang Kun · Journal of Tsinghua University(Science and Technology) · 2008
A channel error resilience algorithm developed for vocoder line spectrum pairs(LSP) parameters based on the source and channel properties enhances the reconstructed voice quality for voice transmission on wireless time-varying,high error prone channels.The vocoder mode information is recovered by a revised maximum posterior transition probabilities algorithm based on long term statistics.The LSP parameters are then mode-separated recovered using the source and channel information.The algorithm requires no extra bandwidth or algorithm delay while greatly reducing the spectrum distortion and enhancing the error resilience ability of the vocoder and the reconstructed voice quality for severe channel conditions.Simulations show that the spectrum distortion is reduced by 0.3 and the mean opinion score(MOS) is improved by about 15%.