Reduced complexity algorithm for channel optimized switched split vector quantization: Application to wideband LSF parameters

In this paper, a reduced complexity stochastic joint source-channel coding system is proposed for efficient and robust coding of wideband speech LSF (Line Spectral Frequencies) parameters. The aim of the initial version of this system was to achieve a transparent quantization quality of LSF parameters under noiseless channel conditions. It was developed based on the switched split vector quantization (SSVQ) technique and called “LSF-SSVQ coder”. After that, we were interested on the improvement of the LSF-SSVQ coder robustness for real transmissions over a noisy channel. To protect implicitly our wideband LSF encoders, we developed a non-redundant joint source-channel coding system based on a reduced complexity stochastic version of the channel optimized SSVQ technique.