HUMAN ACTIVITY RECOGNITION BY USING FUZZY ROUGH PROTOTYPE SELECTION by HUANG RONG HUI Thesis Supervisor: Prof. Simon James, Fong In the past decades, with more diversifications and fewer limitations in the research field, human activity recognition (HAR) has become an active research subject. Thus, human physical activity recognition is now a compelling application in surveillance, human behavior modeling, sports coaching and human computer interaction (HCI), etc. Human activity recognition contains various state-of-the-art methods and techniques. However, physical activities classification is one of the mainstream research fields which presents active in recent years. In this paper, the author proposed a novel Prototype Selection method called Fuzzy Rough Prototype Selection (FRPS), based on fuzzy rough set theory for human activity recognition. In this experiment, a comparison of FRPS with other Prototype Selection methods has proven that the proposed method performs well. FRPS is an effective way in detecting the actual misclassified instances and improved the performance of classification accuracy based on KNN rule.