EPIC: A System Framework for Efficient Egocentric Perception on Embodied AR Glasses

2026-06-14

AI Summary

This work addresses the severe power and memory challenges faced by intelligent AR glasses due to continuous processing of high-resolution first-person video. To overcome these limitations, the authors propose EPIC, a system framework that, for the first time, leverages lightweight multimodal signals—including gaze, head pose, and inertial measurement unit (IMU) data—within an algorithm-hardware co-design approach to dynamically infer user intent. Based on this inference, EPIC selectively retains only the most critical regions of the perceptual stream. This strategy achieves comparable accuracy on video understanding tasks while drastically reducing resource consumption, yielding an average 27.5× reduction in memory footprint and a 24.3× decrease in energy usage compared to processing the full video stream.

Abstract

Modern smart AR glasses are evolving into intelligent systems that support foundation model-based assistance through continuous perception of the user and surrounding environment. However, this perception-first design creates major bottlenecks. Continuously capturing, processing, and storing rich perceptual streams, especially high-resolution egocentric video, imposes substantial power and memory overhead, which is difficult to sustain on resource-constrained AR glasses. In this work, we propose EPIC, an efficient egocentric perception system for embodied intelligence on smart AR glasses. EPIC is an algorithm-hardware co-optimization framework that leverages gaze, pose, and inertial signals to infer user intent and retain only the most informative parts of high-resolution perceptual input, greatly reducing perception overhead. Our results show that EPIC reduces memory footprint by $27.5\times$ and energy consumption by $24.3\times$ on average compared with full video baseline solution, while preserving intelligent assistance accuracy on egocentric video understanding tasks, a key application scenario for embodied intelligence on smart glasses.

Problem

• egocentric perception
• AR glasses
• resource constraints
• perceptual overhead
• embodied intelligence

Innovation

• egocentric perception
• algorithm-hardware co-optimization
• user intent inference
• embodied intelligence
• AR glasses