Investigating Encoding and Perspective for Augmented Reality

TL;DR

This study addresses how encoding (path, object, line, human) and perspective (1P vs 3P) influence AR-based motion guidance across three isolation exercises with varying visibility. Using a 4×2 within-subject design in an HMD/AR setup, it demonstrates that there is no one-size-fits-all encoding or perspective: 3P tends to improve usability for lower-visibility movements, while path and line cues do not consistently outperform simpler target-based cues. The work provides empirically grounded guidelines for deploying immersive visualizations in AR to support rehabilitation and training, emphasizing alignment of design choices with movement visibility and user goals. These findings advance practical AR guidance by clarifying when additional visual information aids or hinders precise motion execution and user experience.

Abstract

Augmented reality (AR) offers promising opportunities to support movement-based activities, such as personal training or physical therapy, with real-time, spatially-situated visual cues. While many approaches leverage AR to guide motion, existing design guidelines focus on simple, upper-body movements within the user's field of view. We lack evidence-based design recommendations for guiding more diverse scenarios involving movements with varying levels of visibility and direction. We conducted an experiment to investigate how different visual encodings and perspectives affect motion guidance performance and usability, using three exercises that varied in visibility and planes of motion. Our findings reveal significant differences in preference and performance across designs. Notably, the best perspective varied depending on motion visibility and showing more information about the overall motion did not necessarily improve motion execution. We provide empirically-grounded guidelines for designing immersive, interactive visualizations for motion guidance to support more effective AR systems.

Figures (7)

• Figure 1: Schematic of the tested visualizations for shoulder flexion. Each visualization has an encoding---a) path, b) object, c) line, and d) human---and perspective---1) first-person and 2) third-person. Examples of object in first person (e) and third person (f) viewed from the HMD.
• Figure 2: (1) At study start, participants held four poses to capture coordinates for shoulder flexion (b), trunk rotation (c), and hip abduction (d). These exercises occur in different cardinal planes (2): sagittal, transverse, and coronal.
• Figure 3: Results from our ANOVA and Kruskal-Wallis tests. Bolded text indicates significant findings ($p<0.05$), standard text indicates marginal effects ($p<0.1$), and grey text indicates non-significant effects.
• Figure 4: Means and $95\%$ confidence intervals for DTW path error, real-time usability ratings (1--5, where 5 is the most usable), and post-survey usability rankings (1-8, where 1 is the most usable). * indicates significance with $p < 0.05$, ** indicates $p<0.01$, and *** indicates $p<0.001$.
• Figure 5: Ranking comparison of visualization techniques for shoulder flexion and hip abduction exercises by ease of use and anticipated frequency (lower ranking = higher usability). First-person (1P, darker colors) and third-person (3P, lighter colors) perspectives are shown for human, line, object, and path visualizations. 3P and object visualizations consistently received better usability rankings.