- Development of a wearable biofeedback system for real-time gait rehabilitation in post-stroke patients using inertial measurement units and machine learning.

Project Abstract

This study presents the design, implementation, and evaluation of an innovative wearable biofeedback system for real-time gait rehabilitation in post-stroke patients, integrating inertial measurement units (IMUs) and machine learning to deliver personalized, adaptive therapy. The system comprises a lightweight wearable sensor array positioned on the lower limbs and trunk, a cloud-enabled processing module for real-time data fusion, and a user-facing feedback interface that provides multimodal cues (visual, auditory, and haptic) to guide corrective gait patterns during therapy sessions. The core aim is to improve gait symmetry, reduce compensatory movements, and accelerate motor recovery by delivering timely, task-specific feedback that reinforces correct motor commands and facilitates motor learning. We developed a robust data collection protocol involving stroke survivors and age-matched controls performing standardized gait tasks across controlled and real-world environments. The IMUs captured accelerometer, gyroscope, and magnetometer data at high sampling rates, from which features such as spatiotemporal parameters, kinematics, variability measures, and asymmetry indices were extracted. A supervised machine learning framework was trained to distinguish between compensatory and normative gait patterns, using a combination of classical machine learning classifiers and deep learning models to ensure high accuracy, low latency, and interpretability. The research emphasizes clinical relevance by incorporating patient-specific baselines and progression tracking, allowing the system to adapt feedback thresholds as rehabilitation progresses. A pivotal component of the work is the real-time feedback engine, which translates algorithmic decisions into actionable cues. Visual cues (e.g., live gait graphs and color-coded indicators), auditory prompts (e.g., cadence and step-length reminders), and haptic signals (e.g., subtle vibration for stance-stance timing adjustments) are synchronized with the gait cycle to minimize cognitive load and enhance motor relearning. The study also investigates system usability and acceptance among both patients and clinicians, employing standardized scales and qualitative interviews to assess perceived benefit, ease of use, comfort, and integration into existing rehabilitation workflows. To validate effectiveness, a randomized controlled pilot trial compared the biofeedback system against conventional gait therapy. Primary outcomes included improvements in gait symmetry (step-length and stance-time asymmetry), velocity, and endurance (6-minute walk test), while secondary outcomes encompassed functional mobility, balance confidence, and quality of life. Statistical analyses examined within-subject improvements, between-group differences, and dose-response relationships, with attention to subgroup effects based on stroke severity and time since onset. The results demonstrate that real-time, IMU-driven biofeedback significantly enhances gait rehabilitation outcomes, yielding greater improvements in symmetry and functional mobility than standard therapy and sustaining gains at follow-up. The system exhibited favorable usability metrics and robust performance in diverse environments, with low computational latency and acceptable battery life. This work advances accessible, data-driven rehabilitation by enabling scalable, personalized gait therapy outside traditional clinical settings, potentially reducing long-term disability and healthcare costs. Future work will focus on expanding multimodal feedback, refining personalization algorithms, and conducting larger multicenter trials to establish generalizability across heterogeneous stroke populations.

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. Identify gait patterns in post-stroke patients using simple wearable sensors.
2. Develop real-time feedback that helps users adjust their walking in a safe, intuitive way.
3. Apply machine learning to distinguish healthy versus impaired gait and trigger feedback appropriately.
4. Test system usability with volunteer participants and gather user feedback.
5. Evaluate short-term improvements in walking symmetry and confidence.

---

---

What You Will Do Step by Step

---

1. Learn basics of gait and stroke-related impairments; review safety requirements.
2. Choose affordable wearable sensors (e.g., small sensors placed on leg).
3. Collect simple walking data from volunteers, recording timing and movement cues.
4. Process data to extract easy-to-understand measures of gait (e.g., step length, cadence).
5. Develop a straightforward feedback system (audio or haptic) that guides improvements in real time.
6. Train a simple machine-learning model to recognize problematic gait patterns.
7. Evaluate the system in short walking sessions and record user experiences.

---

---

Expected Outcome

Expect a usable wearable system that provides real-time feedback to help post-stroke patients walk more steadily, supporting faster adaptation and greater confidence in daily activities.