Unsupervised clustering for abnormality detection based on the tri-axial accelerometer

Today´s society confronts the aged problem by the reason that the rate of the population over age 65 increases rapidly in world wide. One important problem is how to manage them effectively from the dangerous emergency like falling, slipping or unintended activity. Such activity is required to be detected as the abnormal behavior to predict the dangerous emergency of elderly people so that they are protected from more fatal situation. There are many researches to classify the behavior but it is evaluated that it is not proper to detect just only the abnormal behavior. We propose the method of unsupervised learning to overcome the disadvantage of the supervised learning method only for abnormal activity detection. In experiment, we show that the unsupervised learning can be used to detect abnormal behavior from three subjects.