Spherical World-Locking for Audio-Visual Localization in Egocentric Videos
This addresses spatial synchronization issues in egocentric video understanding for applications in user and scene analysis, representing an incremental improvement over conventional head-locked methods.
The paper tackles the problem of spatial synchronization in egocentric videos by proposing Spherical World-Locking (SWL), a framework that transforms multisensory streams based on head orientation to offset self-motion challenges, resulting in improved performance on tasks like audio-visual active speaker localization and auditory spherical source localization.
Egocentric videos provide comprehensive contexts for user and scene understanding, spanning multisensory perception to behavioral interaction. We propose Spherical World-Locking (SWL) as a general framework for egocentric scene representation, which implicitly transforms multisensory streams with respect to measurements of head orientation. Compared to conventional head-locked egocentric representations with a 2D planar field-of-view, SWL effectively offsets challenges posed by self-motion, allowing for improved spatial synchronization between input modalities. Using a set of multisensory embeddings on a worldlocked sphere, we design a unified encoder-decoder transformer architecture that preserves the spherical structure of the scene representation, without requiring expensive projections between image and world coordinate systems. We evaluate the effectiveness of the proposed framework on multiple benchmark tasks for egocentric video understanding, including audio-visual active speaker localization, auditory spherical source localization, and behavior anticipation in everyday activities.