AI-Powered Wearable System for Early Prediction of Freezing of Gait (FoG) in Parkinson's Disease
- 🧠 AI-Powered Healthcare System
- 🦿 Wearable IoT Prototype
- 📱 Flutter Mobile Application
- 🤖 Machine Learning & Deep Learning
- 📊 90.73% Classification Accuracy
- 🎯 Early Prediction of Freezing of Gait (FoG)
- 📑 Research Poster Successfully Presented & Defended
- 🏛️ Military Technical College (MTC)
- 🎓 Bachelor Graduation Project (2025/2026)
- Overview
- Key Milestones
- Project Objectives
- Key Features
- Problem Statement
- Solution
- System Architecture
- Hardware Prototype
- AI Pipeline
- Machine Learning Models
- Deep Learning Model
- Mobile Application
- Datasets
- Results
- Documentation
- Gallery
- Demonstration
- Future Work
- Team
- License
The research poster was officially accepted, presented, and successfully defended during the graduation poster evaluation session at the Military Technical College (MTC), Egypt.
Developed as a Bachelor Graduation Project (2025/2026) within the College of Computing and Information Technology.
Gait Guard is an intelligent AI-powered wearable healthcare system developed to predict Freezing of Gait (FoG) episodes before they occur in patients with Parkinson's Disease.
The project integrates Artificial Intelligence, Wearable IoT Hardware, Embedded Systems, Motion Analysis, and a Flutter Mobile Application into one unified healthcare platform capable of continuously monitoring patient gait and providing early warning alerts.
Unlike conventional systems that detect FoG after it occurs, Gait Guard focuses on early prediction, allowing patients to react before the freezing episode begins, helping reduce falls and improve mobility.
- Early prediction of Freezing of Gait (FoG)
- Reduce fall risk for Parkinson's patients
- Continuous gait monitoring
- Real-time wearable sensing
- AI-assisted decision making
- Mobile healthcare monitoring
- Affordable wearable solution
- Improve patient quality of life
- ✅ Early FoG Prediction
- ✅ Wearable Ankle Device
- ✅ CNN + BiLSTM + Attention Network
- ✅ Machine Learning Baseline Models
- ✅ Flutter Mobile Application
- ✅ ESP32-S3 Embedded Platform
- ✅ MPU6050 Motion Sensor
- ✅ Real-Time Monitoring
- ✅ Intelligent Alert System
- ✅ Healthcare AI
- ✅ Embedded IoT Solution
- ✅ Motion Signal Processing
- ✅ Research-Based Development
- ✅ Mobile Visualization
- ✅ Patient Safety Enhancement
| Category | Technologies |
|---|---|
| Programming | Python |
| Machine Learning | Scikit-learn |
| Deep Learning | TensorFlow, CNN, BiLSTM, Attention |
| Mobile Development | Flutter |
| Hardware | ESP32-S3, MPU6050 |
| Design | Figma |
| Documentation | PDF, Research Paper, Poster |
Parkinson's Disease affects millions of people worldwide and is often accompanied by Freezing of Gait (FoG), one of its most disabling motor symptoms.
During a FoG episode, patients suddenly lose the ability to continue walking despite intending to move, significantly increasing the likelihood of falls and severe injuries.
Existing monitoring systems primarily detect FoG after the event has already started, limiting their effectiveness in preventing accidents.
Our objective was to design an intelligent wearable solution capable of predicting FoG before it occurs, enabling timely alerts and improving patient safety.
Gait Guard combines multiple technologies into a single intelligent healthcare platform.
The system includes:
- Wearable ankle-mounted sensing device
- IMU motion data acquisition
- Machine Learning prediction
- Deep Learning prediction
- Mobile application
- Intelligent alert generation
- Continuous gait monitoring
The wearable prototype continuously captures motion data, processes gait information using AI models, and communicates predictions to the mobile application for real-time visualization and patient notification.
Gait Guard follows a complete end-to-end intelligent healthcare pipeline that combines wearable sensing, Artificial Intelligence, and mobile technologies.
- Motion data acquisition using the wearable IMU sensor.
- Signal preprocessing and filtering.
- Window segmentation.
- Feature extraction.
- Machine Learning prediction.
- Deep Learning prediction.
- Decision generation.
- Mobile application visualization.
- Alert notification to the patient.
The wearable prototype was designed to be comfortably mounted around the patient's ankle.
| Component | Purpose |
|---|---|
| ESP32-S3 | Main Microcontroller |
| MPU6050 | Motion Sensor (Accelerometer & Gyroscope) |
| LiPo Battery | Portable Power Supply |
| Buzzer | Audible Warning |
| Vibration Motor | Silent Alert |
| 3D Printed Enclosure | Wearable Housing |
- Lightweight wearable design
- Low power consumption
- Real-time motion acquisition
- Wireless communication
- Comfortable ankle mounting
The AI pipeline consists of several stages:
IMU signals are continuously collected from the wearable sensor.
- Noise removal
- Signal normalization
- Data cleaning
The continuous gait signals are divided into fixed windows suitable for prediction.
Statistical and temporal gait features are extracted for Machine Learning models.
Traditional ML algorithms analyze gait characteristics.
The neural network automatically learns temporal gait patterns.
The prediction result is transmitted to the mobile application.
The following baseline models were implemented and evaluated:
- Random Forest
- Support Vector Machine (SVM)
- Logistic Regression
- Decision Tree
These models provide strong baselines for comparison against the deep learning architecture.
Our deep learning model combines multiple neural network layers to capture both spatial and temporal gait characteristics.
- CNN
- BiLSTM
- Attention Mechanism
CNN extracts spatial features from IMU signals.
BiLSTM learns long-term temporal dependencies.
Attention focuses on the most informative gait segments before prediction.
The Flutter mobile application acts as the user interface of the entire healthcare system.
- Patient Monitoring
- Sensor Connection
- Real-Time Prediction
- Alert Notifications
- Prediction Visualization
- User-Friendly Interface
- Healthcare Dashboard
The application receives prediction results and provides visual feedback for continuous patient monitoring.
The project utilizes publicly available Parkinson's Disease gait datasets.
- Daphnet Freezing of Gait Dataset
- Figshare IMU Dataset
These datasets contain synchronized inertial sensor recordings collected from Parkinson's Disease patients during walking experiments.
| Metric | Value |
|---|---|
| Accuracy | 90.73% |
| Weighted F1-Score | 90.53% |
The proposed system achieved promising performance for early FoG prediction while maintaining balanced classification capability.
The confusion matrix demonstrates strong classification performance across the evaluated classes.
The repository includes complete academic documentation.
- 📄 Documentation.pdf
- 📑 Research Paper.pdf
- 🖼 Poster.pdf
These documents provide detailed explanations of the methodology, implementation, experiments, and evaluation process.
The research poster was officially presented and defended during the Military Technical College graduation poster evaluation.
This repository includes demonstration videos illustrating the complete system.
- Flutter User Interface
- Real-Time Prediction Visualization
▶️ Watch the Mobile App Demo:
https://drive.google.com/file/d/103NYJ6lQ_1Kv16tt1N-uMBQfVKYp13DA/view?usp=drive_link
- Wearable Device
- AI Prediction Workflow
- Mobile Integration
▶️ Watch the System Demo:
https://drive.google.com/file/d/16TMlPehb_BHR3q7B5x8WEalUOmKROm87/view?usp=drive_link
This repository is intended to showcase the project's architecture, methodology, wearable hardware prototype, research documentation, and achieved results.
Certain implementation details, source code, trained model weights, and proprietary assets are not publicly available due to academic research policies and intellectual property considerations.
Future versions of Gait Guard may include:
- Cloud Synchronization
- BLE Communication
- TinyML Deployment
- Edge AI Inference
- Model Compression
- Clinical Validation
- Smartwatch Integration
- Remote Patient Monitoring
- Continuous Health Analytics
- George Reda Lotfy
- Abanoub Magdy Rizk
- Flora Osama Shokry
- Joy Joseph Kamel
- Enjy Boushra Tawfik
- Kholoud Ashraf Ebrahem
- Samuel Remon Gerges
Prof. Nashwa Mahmoud Mohamed El-Bendary
Institution
Arab Academy for Science, Technology and Maritime Transport (AASTMT)
Faculty
College of Computing and Information Technology
Campus
South Valley Campus
Project
Bachelor Graduation Project
Academic Year
2025/2026
- AI-Powered Healthcare System
- Wearable IoT Device
- Embedded AI Solution
- Flutter Mobile Application
- ESP32-S3 Wearable Prototype
- CNN + BiLSTM + Attention Network
- 90.73% Classification Accuracy
- Research Poster Successfully Presented & Defended
- Military Technical College (MTC)
- Graduation Project (AASTMT)
We sincerely express our gratitude to our supervisor, Prof. Nashwa Mahmoud Mohamed El-Bendary, for her continuous guidance and support throughout the development of this project.
We also thank the faculty members of the Arab Academy for Science, Technology and Maritime Transport and the evaluators at the Military Technical College for their valuable feedback during the graduation project assessment.
Finally, we appreciate every team member whose dedication and collaboration contributed to the successful completion of Gait Guard.
This repository is released for educational and portfolio purposes.
© 2026 Gait Guard Team. All Rights Reserved.
Made with ❤️ by the Gait Guard Team