Low complexity tonality control in the Intelligent Gap Filling tool

Abstract

When coding audio signals at low bitrates with a transform coder the most prominent artifacts are spectral holes resulting from spectral lines being quantized to zero. State of the art codecs circumvent this by Noise Filling [1] and Bandwidth Extension (BWE) [2]. Both methods have in common that they do not code parts of the waveform itself but code a coarse description of the signal. At decoder side a synthetic signal is generated and adjusted according to the coded parameters. The presented system called Intelligent Gap Filling (IGF) is a combination of both methods. Spectral holes are filled with random noise or with copied decoded signal components from lower frequency regions. In the latter case a control mechanism is required to adjust the tonality of the copied signal components to reach good audio quality. This paper describes the way of controlling the tonality of IGF. The presented approach is of low complexity and allows for selective application without producing additional algorithmic delay. IGF is part of the 3GPP standard Enhanced Voice Services (EVS) as well as MPEG-H standardized by Moving Picture Experts, Group (MPEG).