HUMAN ACTION RECOGNITION USING SKELETON JOINT IN REAL TIME ENVIRONMENTS

In this paper, we have to propose an effective approach for the Human activity recognition in the real time environment. We recognized several human activities using Kinect. Through the Kinect a 3D skeleton, joints data collected from the real time video in the analogous form of frames and skeleton, joints, orientation, rotation of all the joint angles from the any random selected frames. After extracting the frames we have implemented classification technique PCA (principal component analysis) with some data we have to classify all the activity models, however, we have to conclude the very less number of data (8-12%) to train our system from all the activity frames. After applying the PCA classification techniques we got excellent accuracy 93.6%. Finally, we observe that our proposed techniques are more accurate than other methods; therefore this technique is more suitable in real-time application such as robotics, human computer interface, in game player’s activity recognition.