Enhanced Online Segmentation and Performance Evaluation Method for Real-Time Activity Recognition in Smart Homes

This paper presents a robust method for real-time recognition of Activities of Daily Living (ADLs) in smart home environments using IoT data. Our approach improves the segmentation of sensor data streams into distinct activities by leveraging IoT sensor spatiotemporal features and applies the Needleman-Wunsch method to align predicted and actual activities. Testing on the Aruba dataset achieved 83.2% accuracy, demonstrating superior performance in segmentation and activity recognition compared to existing dynamic methods. Future work will focus on developing a sensor installation simulator to enhance accuracy and reliability.