Results on reverse water-filling, SNR, and log-spectral error in codebook-based coding

This paper identifies optimum levels of reverse water-filling for codebook-based coding of noise and speech signals. We find that there is little to be gained from optimizing an effective rate parameter. We identify trade-offs between SNR and log-spectral error. We show that the use of a gain factor compares favorably with reverse water-filling in some situations.