Activity Recognition from Sensor Data

• Tech Stack: CRF++, Perl
• Algos: Conditional Random Fields (CRF)
• Code URL: View on Github

This project uses conditional random fields (CRF) to classify activities (e.g. walking, sitting, falling down, etc) based on localization or sensor data. This is especially helpful in detecting falls and emergencies that can enhance the safety of elderly individuals living independently. By considering adjacent activities, the method accounts for the sequential nature of the data and achieves over 91% accuracy in activity prediction.