Multichannel audio quality measures enhanced by adaptive feature fusion
Multichannel Audio Quality Assessment: Extending Pretrained Perceptual Models to Spatial Audio
SoundArtificial Intelligence
Summary
Assessing how good spatial audio sounds is hard because quality depends not just on the sound itself but also on how different speakers work together. The authors looked at ways to use existing audio quality tools, originally made for fewer speakers, in setups with many speakers like 5.1 surround sound. They tested different methods and found that combining information at the feature level, which means mixing detailed audio details before the main analysis, worked best. Their method helps measure audio quality more accurately for multichannel sound systems.
What this means in practice
- •For audio engineers: Evaluate and improve the perceived quality of multichannel audio mixes using advanced automated tools for 5.1 surround sound systems.
- •For streaming media platforms: Automatically assess spatial audio quality in multichannel content to optimize streaming delivery and user experience.$Commercial implications: Enables new quality assurance tools for spatial audio streams, improving content reliability for large customer bases.
Authors
Gouthaman KV, Shiv Gehlot, Vishnu Raj, Lars Villemoes, Arijit Biswas
Abstract
Accurate perceptual quality assessment is essential for evaluating and optimizing spatial audio, where perceived quality depends on both signal fidelity and inter-channel spatial relationships. However, subjective evaluation is costly, while existing perceptual models are often trained for limited channel configurations and cannot be directly applied to higher-channel-count audio. This raises the question: how can pretrained perceptual knowledge be effectively reused for multichannel spatial audio? Using 5.1-channel audio, we study four levels of multichannel integration: signal, prediction, latent, and feature and propose two learned approaches: latent-level aggregation of spatial-group representations and the feature-level Feature-Band Group Attention (FGAtt), which adaptively fuses spatial groups at the feature level before perceptual processing. Across five 5.1-channel test sets, FGAtt achieves the strongest over- all performance, demonstrating the effectiveness of feature-level adaptation for reusing pretrained perceptual knowledge