Human Activity Recognition using bag of feature

One of the primary challenges in the identification of human behavior is to identify several human behaviors since identification of various human behaviors is very hard and requires a lot of training data. In this study, using bag-of features containing 5 valuable spatial-temporal features and extracting the most valuable features from features collocation, it has been tried to increase the accuracy of identifying human behavior in videos. The accuracy of the solution proposed in this study was 96% based on the standard database of Cambridge University, KTH which had a favorable capacity compared to other similar solutions. In addition, classification of K NearestNeighbor (KNN) has been used. According to the value of bag-of feature made of human behaviors, the algorithm of classification will have high precession to identify the defined classes.