HPL-2012-190Hand Gesture Recognition from a Single Training Sample
Manjunatha, Varun; Subramanian, Anbumani;
Keyword(s): hand gesture recognition; gesture learning; depth camera; gestural interaction
Abstract: Camera based gesture recognition systems suffer from the severe limitation that an end-user is unable to add any new gestures to the pre-existing database of gestures, due to the rigors of training the database. In this work, we propose and have built, a gesture recognition system which enables the addition of new single-stroke gestures by performing each new gesture only once, thus allowing customization of the gesture recognition system. We tested our method on the Kinect depth camera for a variety of single-stroke hand gestures, including alphanumeric characters and commonly used symbols. We tested our recognition method on several users and observed an average accuracy of 77.6% from a set of 500 samples.
External Posting Date: September 6, 2012 [Abstract Only]. Approved for External Publication - External Copyright Consideration
Internal Posting Date: September 6, 2012 [Fulltext]