A multiple description speech coder based on AMR-WB

The robust speech coder is needed by speech communication over unreliable channel such as internet. A novel multiple description (MD) speech coder based on AMR-WB coder is proposed, which divides the parameters of AMR-WB into two descriptions. The parameters of fixed codebook are divided by interleaving algorithm, and the method to recover the adaptive codebook gain and pitch lag is intra-frame interpolation. The experimental results show that this coder can provide better speech quality than AMR-WB and other existent MD parametric speech coders based on CELP structure at the packet loss rate from 5% to 30%. So this codec is suitable to transfer speech signal in an unreliable and band limited channel.