Performance analysis of human gesture recognition techniques

Today, a lot of research is being done in the area of human gesture recognition due to its various uses such as traffic management, surveillance, healthcare management etc. In this paper the performance of different filters, namely Gabor and Canny for edge detection during preprocessing of image sequences has been studied. Subsequently, the preprocessed images are used for human gesture recognition. The focus is mainly on two gestures-walk and bend. The different classifiers that are used are KNN (K-Nearest Neighbour), NN (Nearest Neighbour) and SVM (Support Vector Machine). The results are then compared for different training dataset sizes for each model. It is found that in general, the Gabor filter gave better results than the Canny edge detection method.