Back to Project List
Sensor-Based Human Action Recognition
2019•✅ Completed
SW action Recognition
AI Flow
Project Overview
Sponsored by MOST Taiwan
Problem
Traditional human activity monitoring relied on manual observation and lacked real-time wearable-based systems for detecting human actions and abnormal conditions such as falling or drowning.
Approach
Developed a sensor-based human action recognition system using smartphone and smartwatch sensor data with Deep Bi-LSTM models and Android/Wear OS applications.
Key Results & Impact
- ✓Delivered a real-time wearable action recognition platform capable of detecting activities such as walking, standing, jumping, falling, and drowning with deep learning-based prediction.
- ✓Implemented in Intelligent Swimming Detection and Tracking Project
- ✓Implemented in published paper “Person Tracking by Fusing Posture Data from UAV Video and Wearable Sensors”
Technologies Used
WearablesIMU SensorsDeep LearningIoT
Additional Information
CategoryApplied Machine Learning / AI
Year2019
StatusCompleted
Related Projects
Intelligent Sootblower
Optimization of Sootblower Operation Using Machine Learning for Reducing Sootblower Steam Consumption
Multi-task AI Model for Swimmer Tracking and Action Recognition
Design And Implementation Of A Lightweight Multi-task Deep Neural Network For Swimmer Action Recognition And Re-identification
Product Matching - Shopee National Data Science Competition 2020
Built a Siamese Neural Network for Shopee product matching and achieved Rank 6/139 in the 2020 national data science competition.