A Wearable Mixed Reality Platform to Augment Overground Walking: A Feasibility Study

Abstract

Humans routinely modify their walking speed to adapt to functional goals and physical demands. However, damage to the central nervous system (CNS) often results in abnormal modulation of walking speed and increased risk of falls. There is considerable interest in treatment modalities that can provide safe and salient training opportunities, feedback about walking performance, and that may augment less reliable sensory feedback within the CNS after injury or disease. Fully immersive virtual reality technologies show benefits in boosting training-related gains in walking performance; however, they lack views of the real world that may limit functional carryover. Augmented reality and mixed reality head-mount displays (MR-HMD) provide partially immersive environments to extend the virtual reality benefits of interacting with virtual objects but within an unobstructed view of the real world. Despite this potential advantage, the feasibility of using MR-HMD visual feedback to promote goal-directed changes in overground walking speed remains unclear. Thus, we developed and evaluated a novel mixed reality application using the Microsoft HoloLens MR-HMD that provided real-time walking speed targets and augmented visual feedback during overground walking. We tested the application in a group of adults not living with disability and examined if they could use the targets and visual feedback to walk at 85%, 100%, and 115% of each individual's self-selected speed. We examined whether individuals were able to meet each target gait speed and explored differences in accuracy across repeated trials and at the different speeds. Additionally, given the importance of task-specificity to therapeutic interventions, we examined if walking speed adjustment strategies were consistent with those observed during usual overground walking, and if walking with the MR-HMD resulted in increased variability in gait parameters. Overall, participants matched their overground walking speed to the target speed of the MR-HMD visual feedback conditions (all p-values > 0.05). The percent inaccuracy was approximately 5% across all speed matching conditions and remained consistent across walking trials after the first overall walking trial. Walking with the MR-HMD did not result in more variability in walking speed, however, we observed more variability in stride length and time when walking with feedback from the MR-HMD compared to walking without feedback. The findings offer support for mixed reality-based visual feedback as a method to provoke goal-specific changes in overground walking behavior. Further studies are necessary to determine the clinical safety and efficacy of this MR-HMD technology to provide extrinsic sensory feedback in combination with traditional treatments in rehabilitation.

Keywords: kinematics; mixed reality; motor learning and control; rehabilitation; visual feedback; walking.

Figure 1

Visualization of the HoloLens MR-HMD. The image on the left (A) shows the vertical and horizontal field of view for rendered holographic objects. The visual feedback (B) consisted of a floating holographic ball in different sections based on distance from the HMD. When maintaining the target speed, the ball remains between 1.5 m and 0.85 m from the subject and is colored green. If the subject walked too slowly and the ball was beyond 1.5 m its color changed to blue providing additional feedback that the user should walk faster. The ball clipped when it was less than 0.85 m from the HMD indicating that the subject should walk more slowly.

Figure 2

Walking speed inaccuracy percentage by trial number. The first condition using visual feedback is indicated in blue, the second tested condition in red, and third tested condition in green. Error bars denote 95% confidence intervals. The first overall trial regardless of speed condition had a larger percent inaccuracy compared with nearly all subsequent walking trials. The within-trial inaccuracy was 10.5% higher (CI95: 7.4, 13.5) in first trial of the first speed condition compared to subsequent trials. We found no significant differences in walking speed inaccuracy between the second and 10th trial of the first speed condition, or the 1st and 10th trial in the subsequent speed conditions. As the first trial overall appears to be an outlier due to a learning effect, we excluded that trial for the subsequent analyses.

Figure 3

Percent difference in walking speed: actual vs. target. A one-sample Wilcoxon signed-rank test found that actual walking speeds of the participants did not differ from the expected speeds during the SS₈₅ (median difference from baseline: −13.8%, z = 0.63, p = 0.53), SS₁₀₀ (median difference from 1.2%, z = 1.23, p = 0.23), and SS₁₁₅ (median difference: 14.2%, z = 0.16, p = 0.88) conditions.

Figure 4

Coefficient of variation in walking speed, stride length, and stride time with and without augmented feedback. (A) Comparison of the baseline SS with SS₁₀₀ condition showed that variability of walking speed did not differ when walking overground with and without MR-HMD feedback (CoV difference: 1.44, CI95: −0.54, 3.42, p = 0.16). (B) There was more variability in stride length (CoV difference: 1.47, CI95: 0.16, 2.78, p = 0.03) and (C) stride time (CoV difference: 1.00, CI95: 0.10, 1.88, p = 0.03) with MR-HMD feedback suggesting that individuals adopted more variable control strategies to achieve the target speed.