Improvement of tandemless transcoding from AMR to EVRC

The simplest way to transcode two different speech coding standards is by the tandem method, but this results in quality loss, computational increase and extra delay. This paper proposes a tandemless transcoding method for Adaptive Multi Rate (AMR)[1] which is the speech coding standard of the asynchronous IMT-2000 system and Enhanced Variable Rate Codec (EVRC)[2]. The proposed algorithm is composed of three parts : (1) LSP conversion using linear interpolation , (2) pitch conversion using excitation signal and (3) adaptive / fixed codebook gains conversion. The proposed algorithm reduces not only processing delay but introduces less degradation to speech than that of the tandem method, which was verified by an objective measure.