Multichannel Audio Quality Assessment: Extending Pretrained Perceptual Models to Spatial Audio

2026-09-29

AI Summary

This study addresses the challenge that pretrained perceptual models struggle to directly evaluate multichannel spatial audio quality. To this end, it explores multichannel integration strategies across four hierarchical levels—from signal to feature—and proposes a feature-level Frequency-band Group Attention (FGAtt) mechanism. By combining latent variable aggregation with multichannel signal processing, the method adaptively fuses spatial group features, enabling effective transfer of pretrained knowledge to 5.1-channel configurations. Experimental results demonstrate that the proposed model achieves state-of-the-art overall performance across five test sets, validating the effectiveness of the feature-level adaptation strategy for multichannel audio quality assessment.

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

Problem

• Multichannel Audio
• Quality Assessment
• Spatial Audio
• Perceptual Models

Innovation

• Multichannel Audio Quality Assessment
• Spatial Audio
• Pretrained Perceptual Models
• Feature-Band Group Attention
• Latent-level Aggregation