Abstract
The use of augmented reality (AR) head-mounted displays is becoming increasingly prevalent due to their ability to provide interactive experiences with high degrees of freedom. However, AR interactions have been shown to be associated with increased biomechanical loads in the neck and shoulders. This repeated-measures study evaluated the effects of interaction errors on neck and shoulder biomechanical loads during AR tasks. Twenty participants performed two standardized AR tasks (omni-directional pointing and cube placing) with and without AR interaction errors. During the tasks, neck and shoulder angles and muscle activity were collected. The results showed that the presence of errors led to increased neck extension and shoulder flexion angles. Peak muscle activity in the shoulders (anterior and medial deltoids) also increased with errors. These findings highlight the importance of reducing interaction errors in AR interfaces to minimize risks of musculoskeletal discomfort and injuries in the neck and shoulders
Keywords
Augmented reality (AR) interactions with head-mounted displays (HMDs) can lead to discomfort in the neck and shoulders because of the additional weight of the HMDs and shoulder movements without adequate upper-limb support (Kim et al., 2020). Additionally, users may experience more interaction errors due to its novel interfaces, technological constraints, and their lack of familiarity with AR interfaces (Alzahrani, 2020). Such errors are known to be associated with increased muscle activity and fatigue (Mehta & Agnew, 2011). Although prior studies indicated that AR interaction errors may increase perceived physical demand (Kia et al., 2021), there remains a gap in research that objectively measures the impact of those errors on biomechanical loads. Therefore, this study aimed to assess the effects of AR interaction errors on the biomechanical loads in the neck and shoulders.
Twenty healthy participants were recruited with an equal gender distribution (10 males and 10 females). The mean (standard deviation) of their age, weight, and height was 26.3 (6.2) years, 68.4 (12.1) kg, and 171.7 (8.5) cm, respectively. Sixteen of them were right-handed and 4 participants were left-handed. In this repeated-measures laboratory study, each participant performed two AR tasks (omni-directional pointing and cube placing) using an AR HMD (HoloLens; Microsoft, Redmond, WA). The AR tasks were adopted from a computer-human interaction study (Kia et al., 2021). Each participant repeated the AR tasks with two different error levels: without errors (one air tap to select the targets) and with errors (three air taps required to select the targets). During the tasks, neck and shoulder angles as well as muscle activity, were collected using a motion capture (Motive 2.0; Optitrack; Natural Point, OR) and a wireless electromyography system (EMG) (WBA; Mega Electronics; Kuopio, Finland). The joint angle outcome measures included neck sagittal extension and shoulder sagittal flexion angles. The muscle activity measures included the right-side splenius capitis, upper trapezius, anterior deltoid, and middle deltoid. The joint angle and muscle activity data were summarized by the 10th, 50th, and 90th percentile values. Generalized linear mixed models (GLMM) were used to test our hypothesis that the error level (independent variable) would affect neck and shoulder angles and muscle activity (dependent variables). The ‘error level’ and ‘participant’ were included as the fixed and random effects, respectively. An alpha level of 0.05 was set for the statistical significance threshold.
The results showed that the neck extension angles during AR interaction with errors (30.8°) were significantly higher than the neck extension during the error-free interaction (10.4°) (p < 0.01). Similarly, the shoulder flexion with errors (60.9°) was higher than the shoulder flexion without errors (53.1°) (p < 0.06). The greater neck and shoulder angles with the errors could also be explained by more repetitive movements and the prolonged, non-neutral neck and shoulder postures from multiple attempts to select the targets. These observed neck and shoulder angles appeared to be large enough to increase the risks of neck and shoulder discomfort and injuries (Bernard, 1997). Therefore, reducing errors in the AR interfaces may be beneficial for reducing risks of neck and shoulder injury.
The muscle activity in splenius capitis, anterior, and medial deltoids tended to be higher by ~3.0 %MVC with AR interaction errors. This trend was in line with the greater neck extension and shoulder flexion angles with errors. While the differences in the muscle activity measures between the AR interaction with and without errors were relatively small (below ~3.0 %MVC), previous studies have shown that even slight increases in muscle activity (as small as 0.5 %MVC) can elevate the likelihood of musculoskeletal discomfort and injuries in the neck and shoulders, particularly for prolonged exertion (Au & Keir, 2007).
In summary, this study showed that the joint angles and muscular load in the neck and shoulders increased in the presence of AR interaction errors. The study findings indicate that minimizing AR interaction errors can reduce the risk of musculoskeletal problems in the neck and shoulder regions.
