Host: The Institute of Image Information and Television Engineers
Generally, electronical slides are used in presentations and meetings. This paper propose a gesture-based marking operation for slide interface, because audiences sometimes miss important parts. Gestures are classified by using HMM (Hidden Markov Model). Some angles of arm joint and hand shape are used as feature value. The experiments are carried out using 3 kinds of HMMs. The first one is Left-to-Right HMM. The second one is combination of Left-to-Right HMM and Ergodic HMM. The last one is weighted combination of Left-to-Right HMM. The best classification performance is achieved by the last HMM with 85% accuracy.
