Model-based deadzone optimization for stack-run audio coding with uniform scalar quantization

In this paper we present a model-based method to optimize the dead- zone of uniform scalar quantization in transform audio coding. The input signal is coded in modified discrete cosine transform (MDCT) domain by uniform scalar quantization followed by context-based arithmetic coding. The optimal deadzone is derived using a non- asymptotic method, assuming that the distribution of MDCT coefficients is approximated by a generalized Gaussian model. We show that deadzone optimization improves slightly quality, especially at low bitrates.