Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This study introduces a framework for analyzing the sensitivity of generative spatial audio metrics in First-Order Ambisonics (FOA) by evaluating how these metrics respond to variations in spatial parameters such as azimuth and elevation. The authors establish three key criteria—Responsiveness, Smoothness, and Symmetry—to assess the performance of various metrics, including Fréchet Audio Distance (FAD) and intensity vectors, across controlled FOA scenes of varying complexity. The results indicate that FAD with localization-specific embeddings and acoustic maps performs well in terms of Responsiveness and maintains robust Smoothness and Symmetry, while intensity vectors show a decline in performance as scene complexity increases.
FAD with localization-specific embeddings outperforms traditional metrics in capturing the nuances of complex spatial audio environments.
Evaluating generative spatial audio for First-Order Ambisonics (FOA) remains challenging due to a limited understanding of how metrics respond to changes in spatial parameters such as azimuth and elevation. We propose a framework to analyze metric sensitivity along continuous spatial trajectories, drawing on principles of sensitivity analysis in parametric sound synthesis. Using controlled FOA scenes with increasing scene complexity, we define three desiderata for metric behavior: Responsiveness, Smoothness, and Symmetry. We assess standard distribution-based and sample-based metrics, including Fr\'echet Audio Distance (FAD), intensity vectors, and acoustic maps. Our findings show that FAD using localization-specific embeddings and acoustic maps yield high Responsiveness and robust Smoothness and Symmetry across conditions, while intensity vectors degrade with increasing scene complexity. This is the first step towards investigating the sensitivity of metrics for generative spatial audio.