SAQAM: Spatial Audio Quality Assessment Metric

TL;DR

SAQAM is a novel deep learning-based metric for assessing spatial audio quality and localizability without subjective data, improving AR/VR sound realism and serving as a differentiable loss function.

Contribution

Introduces SAQAM, a multi-task learning framework that evaluates audio quality and spatialization without human judgments, applicable to enhancing audio processing models.

Findings

SAQAM correlates well with human judgments across datasets.

It is differentiable and improves speech-enhancement networks.

Uses simulated judgments and DOA-based features for assessment.

Abstract

Audio quality assessment is critical for assessing the perceptual realism of sounds. However, the time and expense of obtaining ''gold standard'' human judgments limit the availability of such data. For AR&VR, good perceived sound quality and localizability of sources are among the key elements to ensure complete immersion of the user. Our work introduces SAQAM which uses a multi-task learning framework to assess listening quality (LQ) and spatialization quality (SQ) between any given pair of binaural signals without using any subjective data. We model LQ by training on a simulated dataset of triplet human judgments, and SQ by utilizing activation-level distances from networks trained for direction of arrival (DOA) estimation. We show that SAQAM correlates well with human responses across four diverse datasets. Since it is a deep network, the metric is differentiable, making it suitable as a loss function for other tasks. For example, simply replacing an existing loss with our metric yields improvement in a speech-enhancement network.