Modeling the temporal evolution of LPC parameters for codebook-based speech enhancement

Abstract

Conventional statistical single-channel noise reduction methods suffer from bad performance in highly nonstationary environments. In contrast to that, model-based algorithms have the potential to deal with those adverse conditions. In this paper, wefocus on codebook-basedalgorithms which utilize trained codebooks where typical speech and noise spectral shapes are stored. Speech and noise estimates are determinedframefor frame independentlywhich allows to deal with highly non-stationary noise. Ey incorporating memory, the performance can befurther improved. In this paper, elaborated models for memory modeling are presented and a preliminary validation is provided.