Efficient MDCT-Based Multi-Channel Coding with Perceptual Whitening and Broadband ILD Compensation

Abstract

Stereo and multi-channel audio coding for mobile communication applications is challenging due to bitrate, latency and computational complexity constraints. To address these constraints, a novel MDCT-based stereo coding scheme is introduced. A key feature of this scheme is the broadband inter-channel level compensation of the spectrally and temporally whitened channels, followed by a robust mechanism for selecting between Mid/Side and Left/Right coding for each frequency sub-band. For lower bitrates, inter-channel time and phase difference compensation methods are additionally utilized. Enhancements comprising stereo coding of spectral and temporal noise shaping parameters and stereo-aware bandwidth extension further increase coding efficiency. The paper outlines the fundamental concepts of this MDCT-based stereo coding incorporated into the newly introduced 3GPP IVAS codec and their extension to multi-channel coding. Finally, the IVAS codec is compared to state-of-the-art counterparts, demonstrating its advantages in communication scenarios across a wide range of content types.