From Community Feedback to Measurement Redesign: Iterating mRehab for Accessible Home-Based Stroke Rehabilitation

Abstract

Home-based stroke rehabilitation often suffers from low adherence due to usability barriers and a lack of engaging, trustworthy tools. mRehab addresses these challenges by pairing a smartphone application with 3D-printed functional objects, enabling post-stroke survivors to practice task-oriented, upper-limb activities in their daily environments. To ensure the system meets real-world needs, we conducted a Community Engagement Studio with 22 expert consultants, gathering critical feedback on system accessibility, stroke-specific interaction, clinical integration, and overall trust. These stakeholder insights drove a substantial redesign of mRehab's interaction design and sensing pipeline. To address concerns regarding responsiveness and measurement reliability, we transitioned from legacy fixed moving-average smoothing to a per-axis Kalman filtering approach with baseline-subtracted stillness detection, integrated an RMS-jerk smoothness metric, and revised the physical prop designs and audio instructions. Preliminary trace-level evaluations across four functional activities demonstrate that the new Kalman pipeline effectively eliminates system lag (0 to 17 ms) compared to the previous 25-sample moving average (204 to 221 ms). By translating community feedback into concrete algorithmic and hardware improvements, this work establishes a highly responsive, user-centered foundation for future clinical-effectiveness validations.