LAUSR

In recent years, sensor-based activity recognition has integrated the field of sensor networks with data mining techniques to model a broad range of human activities and behaviour. Huge amounts of sensor data coming from smart gadgets such as smartphones and smartwatches opens up the possibility of probing and extracting knowledge from the data in the direction of monitoring and health care. Due to the immense popularity and extensive use of smart gadgets equipped with sensors, it is more realistic and effective to utilize them in the activity recognition systems. Sensor-based activity recognition is a challenging task due to the heterogeneous nature and noisy aspect of sensor data. This work presents an ontology-based knowledge model that conceptualizes the task of human activity recognition. The knowledge model is based on two newly developed ontologies: Sensor Measurements Ontology to model the heterogeneous sensor data and Activity Recognition Ontology to conceptualise the activity recognition process by capturing the relationships between the low level acts (simple activities) and high level (inferred activity). Besides activity recognition, the proposed model handles the issue of sensor heterogeneity and provides reusability, interoperability and exchange. The proposed model is validated with a real life dataset containing sensor observations of 60 users with more than 300,000 (three hundred thousand) samples to illustrate the functionalities in the task of human activity recognition.