The application of subband coding to improve quality and robustness of the sinusoidal transform coder

Abstract

In an attempt to improve the performance of the sinusoidal transform coder (STC) in the presence of channel errors, a structure evolved that was similar to a subband coder. Using some new ideas for subband coding, such as vector quantization of the subband energies and adaptive bit allocation based on the decoded energies, a new quantization algorithm for STC evolved that was not only more robust in channel errors, but led to improved quality, particularly at the 2400 bit/s data rate. It may be possible further to improve the speech quality at this rate by using vector quantization to code the shape of the subband channel gains.<>