Metadata-assisted spatial audio coding in IVAS codec
Adriana Vasilache, Tapani Pihlajakuja, Mikko-Ville Laitinen
Abstract
The 3GPP Immersive Voice and Audio Services (IVAS) codec enables mobile spatial communication through coding of the metadata-assisted spatial audio (MASA) format. The MASA format is a new parametric audio format designed for spatial audio capture and representation from varying mobile device microphone arrays. MASA format is composed of one or two audio transport channels and associated spatial metadata. The amount of uncompressed data in MASA can be up to almost 2 Mbps including the 422.4 kbps of spatial metadata. As the IVAS codec supports bitrates ranging from 13.2 kbps to 512 kbps, the compression rate goes from as low as 0.7% to 25%. This paper presents the flexible coding flow of the MASA format in IVAS to support the entire bitrate range. It also describes the bitrate allocation between transport channels and metadata, and the techniques employed for reducing the metadata prior to coding while maintaining perceptually representative information. The encoding performance is demonstrated with both objective and subjective measures.
BibTeX
@inproceedings{icassp2025_metadataassisted,
title = {Metadata-assisted spatial audio coding in IVAS codec},
author = {Adriana Vasilache and Tapani Pihlajakuja and Mikko-Ville Laitinen},
booktitle = {ICASSP 2025},
year = {2025}
}