This project develops a machine learning model to automatically recognize and classify human activities using smartphone accelerometer and gyroscope data. The system accurately predicts six fundamental daily activities: walking, walking upstairs, walking downstairs, sitting, standing, and laying down.
- Sensors: Tri-axial accelerometer and gyroscope data from 30 participants
- Features: 561 statistical and frequency domain features extracted from sensor signals
- Processing: Signal filtering, window segmentation, and feature normalization
- Classification: Multi-class supervised learning with SVM, Random Forest, or Neural Networks
- Validation: Cross-validation for robust performance evaluation
- Performance: High accuracy real-time activity classification
Healthcare: Patient monitoring, elderly care, rehabilitation tracking
Fitness: Automated activity logging, calorie estimation, goal monitoring
Smart Technology: Context-aware automation, personal assistants, location services
- Real-time processing with low latency
- Device-independent across smartphone models
- Robust performance across diverse user patterns
- Complete automated pipeline from raw data to predictions
This project demonstrates practical machine learning application for ubiquitous activity recognition using everyday smartphone technology.