A framework for human activity recognition based on accelerometer data

Monitoring and classification of human activity has been an active area of research for the past few years due to the increasing demands in healthcare sector. Quick aid for falls in elderly persons and detecting emergency situations are few leading cause of such interest. In this paper, a human activity recognition system based on motion patterns on a smartphone is proposed for classification of activities such as fall, walk, run, ascending, and descending stairs. The binned distribution based feature of acceleration data has been used for classification purpose. A systematic approach for classification of different activities using threshold and multistage Support Vector Machine (SVM) has been developed. Experimental results show considerable accuracy in activity recognition with the proposed scheme.