Mirror Therapy (MT) effectively supports post-stroke upper-limb rehabilitation but requires professional supervision and often leads to low patient motivation. While digital MT systems address these limitations, they typically compromise the embodiment benefits of traditional mirrors. Through informant-driven design with interdisciplinary experts, we developed A-MIrror, an augmented reality mirror system that preserves the view of both mirrored hand and overlays digital guidance. Using dual cameras for eye and hand tracking, the system creates a visual illusion where both real and mirrored hands appear to naturally interact with virtual 3D objects on screen. Our evaluation with 14 post-stroke patients and 8 therapists demonstrates that A-MIrror significantly enhances motivation compared to traditional MT (p < .01) while achieving comparable embodiment experiences and even superior illusion latency (30.47% faster, p < .001) to traditional mirrors. This study presents a promising approach for independent post-stroke rehabilitation that integrates the strengths of both traditional and digital MT, offering insights for enhancing future digital rehabilitation applications.
ACM CHI Conference on Human Factors in Computing Systems