Development of an AI-driven gait analysis and rehabilitation feedback system using wearable sensors for stroke survivors

Project Abstract

Stroke remains a leading cause of long-term disability worldwide, with impaired gait significantly impacting mobility, independence, and quality of life. This study presents the development and evaluation of an AI-driven gait analysis and rehabilitation feedback system that leverages wearable sensors to monitor, analyze, and guide gait recovery in stroke survivors. A multidisciplinary approach was adopted, integrating biomechanics, machine learning, and human-centered rehabilitation paradigms to create a practical, portable solution suitable for clinical and home settings. The system comprises unobtrusive wearables placed on the lower limbs and trunk, a cloud-based processing pipeline, and an adaptive feedback module delivered through a mobile application and in-clinic display. Gait data, including spatiotemporal parameters, joint angle trajectories, symmetry indices, and temporal sequencing, are extracted in real-time and augmented with contextual information such as fatigue and environmental factors. A supervised learning framework was developed to classify gait abnormality patterns and predict functional recovery trajectories, trained on a dataset collected from a cohort of post-stroke participants across multiple rehabilitation stages. The AI component continuously updates personalized rehabilitation targets, selecting evidence-based interventions such as task-oriented gait training, rhythmic cueing, and strength optimization, and provides real-time feedback to correct asymmetries and improve safety. A modular feedback strategy was implemented, offering corrective cues (visual and haptic), progression of task difficulty, and motivational progress summaries, designed to maximize adherence and motor relearning. The system was validated through a two-phase study a controlled clinical trial with 60 participants comparing standard gait rehabilitation to the AI-assisted program, and a longitudinal home-use study with 20 participants over eight weeks to assess usability, adherence, and transfer of gains to daily life. Outcome measures included standardized gait scales (e.g., Dynamic Gait Index, #10-Meter Walk Test), kinematic symmetry metrics, locomotor endurance, and patient-reported outcomes on confidence and daily activity participation. Results demonstrated significant improvements in gait symmetry, reduce compensatory patterns, and faster initiation and execution of steps in the AI-assisted group versus conventional therapy, with effect sizes indicating clinically meaningful gains. The home-use cohort reported high acceptability, with adherence rates exceeding 85% and positive perceived impact on independence. The AI model showed robust generalization across participants and settings, maintaining high accuracy in event detection and error prediction despite variability in sensor placement and compensatory strategies. Potential limitations include sensor drift, data privacy considerations, and the need for clinician oversight to tailor targets. The study contributes a scalable, ecologically valid framework for integrating wearable-derived gait analytics and AI-driven rehabilitation feedback into stroke care pathways, enabling ongoing monitoring, individualized progression, and enhanced translation of learned strategies to real-world ambulation. Future work will explore multi-modal sensor fusion, integration with telerehabilitation platforms, and randomized trials across diverse stroke populations to establish broader efficacy and cost-effectiveness.

Project Overview

What This Project Is About

---

A plain-language overview of the topic and what the project investigates.

---

---

The Problem It Addresses

---

What problem or gap this project tackles and why it matters to the field or society.

---

---

Objectives of the Project

---

1. Understand how gait patterns change after stroke and what sensors can reveal.
2. Build a simple AI-based system to analyze walking data from wearable sensors.
3. Create feedback that guides users to correct movements during rehabilitation.
4. Evaluate the usability and potential impact for therapists and patients.

---

---

What You Will Do Step by Step

---

1. Review basic stroke rehabilitation concepts and gait basics.
2. Set up or select wearable sensors to collect walking data.
3. Collect data from volunteers or simulate data if needed.
4. Process data to extract easy-to-understand gait metrics (e.g., step length, symmetry).
5. Develop a simple AI model to recognize abnormal patterns and provide feedback cues.
6. Design a user-friendly feedback interface for patients and clinicians.
7. Test the system for accuracy, safety, and usability with a small group.
8. Document findings and propose improvements or future work.

---

---

Expected Outcome

A functional, easy-to-use gait analysis and feedback prototype that can assist stroke survivors in home or clinic rehab, with initial evidence of usefulness and clear next steps for refinement.