Classifying Human Body Postures by a Support Vector Machine with Two Simple Features

Human behaviour analysis helps to monitor a person's daily activities and detect home care emergencies. Classifying posture is an important step of human behaviour analysis. Many studies improve the accuracy of classifying. However, the number of features is big or extracting these features uses complicated formulas. Therefore, we proposed two features with simple computing. Two new features are formulated from the height and the square showing the human body's silhouettes. Then, we choose a non-linear Support Vector Machine to classify postures based on proposed features. Experiments show Support Vector Machine classify effectively and better than other methods.