Audio Coding Using Masking-threshold Adapted Perceptual Filter
Wen‐Whei Chang, Chin-Tung Wang · 2005
Most current work in LPC-based coding employs either postfitering or noise spectral shaping to take into account the auditory masking effect. In both cases, the LPC parameters are used to characterize both the synthesis filter and the perceptual weighting filter. Since they are originally derived to model the vocal tract, LPC parameters can not precisely describe the nonlinear sensitivity of human car to frequency and to loudness. In this paper, we explore the benefits of masking-threshold adapted perceptual filter for excitation search in analysis-by-synthesis predictive coders. Furthermore, one derivation of Bark spectral distortion is also studied as the objective measure for assessng audio quality.