Support vector machine algorithm for human fall recognition kinect-based skeletal data

Falls are the major reason of injury and accidental death for older people. It is important to recognize falls early for assistance and treatment. In this paper, a Support Vector Machine (SVM) algorithm for recognition falls and other activities based on skeletal data is proposed. Skeletal data, which will be extracted from capturing human body using a Kinect camera system, are obtained on three persons. In order to distinguish falling states such as lying, sitting and standing, the SVM will be applied for training and testing to validate the obtained data. There are three experiments were performed to recognize three circumstances of fall and non-fall, fall and standing, fall and sitting. Experimental results show with the high accuracy of recognition activities to illustrate the effectiveness of the proposed approach.