Predicting User Actions in Kinect using Simple Markov Assumption

This paper proposed a combination of gesture recognition system using Kinect camera and user´s action prediction using a Markov model that makes use of user´s pattern of action sequences. Prediction of future action will help the user speed up while doing a task and potentially avoid an error. The system will record the user chain of command and created a transition probability matrix that will calculate the probability of a command appearing after a certain command is executed. The system has the advantage of no training needed and adaptable with the user habit after a certain amount of time. The proposed system was tested with Kinect using an Image Previewing program and proves to be working.