Smartphone-based assistive gait training for post-stroke rehabilitation using real-time feedback and machine learning analysis

Project Abstract

This study presents a smartphone-based assistive gait training system designed to enhance post-stroke rehabilitation through real-time feedback and machine learning analysis. The system integrates a wearable sensor array (inertial measurement units and plantar pressure sensors) with a user-friendly mobile application to deliver personalized gait rehabilitation protocols in home and clinical environments. Real-time data are streamed to the smartphone, where edge-computing algorithms extract spatiotemporal gait features, symmetry indices, and joint kinematics. A machine learning framework, comprising supervised and reinforcement learning models, analyzes performance trajectories, adapts task difficulty, and optimizes feedback strategies to maximize motor learning and adherence. The human–computer interaction component emphasizes intuitive visualization, motivational cues, and safety features, including fall risk alerts and remote clinician monitoring. The study adopts a mixed-methods design with a longitudinal, randomized controlled trial and qualitative usability assessments. Participants with chronic post-stroke hemiparesis are recruited and randomly assigned to either the smartphone-assisted intervention or conventional home-based physiotherapy. Primary outcomes include changes in gait speed, cadence, stride length, and the Rivermead Mobility Index, assessed at baseline, mid-intervention, post-intervention, and follow-up. Secondary outcomes involve gait asymmetry indices, dynamic balance measures, energy expenditure, and quality of life indices. The system’s real-time feedback modalities—visual, auditory, and haptic cues—are tuned via online learning to promote correct motor patterns, reduce compensatory strategies, and sustain motivation. Machine learning components are trained on labeled gait data to classify gait phases, detect deviations from normative patterns, and predict imminent instability, enabling proactive coaching and safety interlocks. Data privacy and security are addressed through on-device processing with encrypted cloud backup and role-based access for clinicians. Sensor fusion techniques enhance robustness to motion artifacts and inter-individual variability. The project also investigates transfer effects to daily activities and community ambulation, evaluating generalizability across different environments and footwear. A process evaluation explores user experiences, adherence determinants, and perceived barriers to long-term use. Statistical analyses employ intention-to-treat principles, mixed-effects modeling to account for repeated measures, and effect size estimation to determine clinically meaningful improvements. The anticipated outcomes include superior improvements in gait rehabilitation metrics and functional mobility for the smartphone-assisted group, accelerated motor learning curves, and enhanced long-term adherence compared with traditional therapy alone. Potential limitations such as device wearability, data privacy concerns, and variability in baseline impairment are addressed through stratified randomization, customizable intervention doses, and iterative UI/UX refinements. Overall, the research aims to demonstrate that mobile sensor fusion, real-time feedback, and adaptive machine learning can transform gait rehabilitation by delivering personalized, scalable, and accessible therapy that complements conventional care.

Project Overview

What This Project Is About

---

The project explores how a smartphone app can guide and improve walking recovery after a stroke. It uses real-time feedback to help users adjust their gait and applies simple machine learning ideas to tailor tips and exercises to individuals.

---

---

The Problem It Addresses

---

---

---

Objectives of the Project

---

1. Identify common gait problems seen after stroke.
2. Develop a smartphone-based system to detect gait issues using the phone’s sensors.
3. Provide real-time feedback and simple exercises tailored to the user.
4. Evaluate how the system affects usability, motivation, and basic gait improvement.
5. Discuss potential for wider adoption in home rehabilitation.

---

---

What You Will Do Step by Step

---

1) Review basic gait science and rehabilitation needs after stroke. 2) Design a smartphone app interface with sensor input and feedback prompts. 3) Collect data from volunteers performing walking tasks. 4) Process sensor data to flag gait irregularities. 5) Implement real-time feedback and home exercise plan. 6) Run a simple user study on usability and preliminary effectiveness. 7) Analyze results and discuss limitations and future work.

---

---

---

---

Expected Outcome