Rozpoznávání lidských činností ve videosekvencích

Abstract

This master thesis is focused on recognizing human actions in videosequences. Overall, there are proposed three different methods for recognition of human actions. The first method, based on the DTW algorithm, serves only for recognizing actions where is begin and end known. Another two methods are created using neural networks and focus on recognizing human actions in real time. For these methods, I created a test application programmed using C++ language and OpenCV open source library. Then I compared the results and explained advantages and disadvantages particular methods.