Residual Lsf Vector Quantization Using Arma Prediction

Abstract

The residual LSF vector quantization yields bit rate reduction in the vocoders. In this work, a residual LSF vector quantization obtained from Auto Regressive Moving Average (ARMA) prediction is proposed for designing codebooks at very low bit rates. This residual quantization method is applied to multi stage vector quantization method and codebooks are designed. For each codebook, the effectiveness and quality are investigated by calculating the spectral distortion and outliers. The proposed quantization method reduced the distortion without any additional complexity.