Hand gesture recognition using self organizing map for Human Computer Interaction

Hand gesture based Human Computer Interaction is one of the most intuitive yet innovative ways of communication between people and machines. This paper puts forward a new approach for hand gesture recognition using Self Organizing Maps (SOM). Instead of extracting features from the image and using them for training and recognition, the image itself is transformed into radial enclosed contour and trained using SOM. A dictionary of 18 gestures from the standard American Sign Language is defined and a database is created from ten subjects on which the system is tested and evaluated. The average efficiency of prediction was found to be around 92%. Moreover, this approach was found to work well in uncontrolled environments.