Consecutive packet loss concealment algorithm based on SELP vocoder

Vocoder quality deteriorates greatly when consecutive packet loss occurs.This paper presents a mode based linear algorithm to predict the characteristic parameters during packet loss.The algorithm utilizes a short-time correlation of the parameters,with the sub-band voiced/unvoiced parameters as the mode information to calculate the time-domain correlation coefficients of each parameter for different modes.The algorithm predicts the characteristic parameters of the lost frame using the parameters of the preceding correctly received frame to reconstruct the lost frame.Simulations show that when the lost packet length ranges between 75 and 200 ms,the algorithm increases the objective mean opinion score(MOS) of the synthesized speech by 0.03 compared to the traditional packet loss concealment method.