Australian sign language recognition using moment invariants

Abstract

Human Computer Interaction is geared towards seamless human machine integration without the need for LCDs, Keyboards or Gloves. Systems have already been developed to react to limited hand gestures especially in gaming and in consumer electronics control. Yet, it is a monumental task in bridging the well-developed sign languages in different parts of the world with a machine to interpret the meaning. One reason is the sheer extent of the vocabulary used in sign language and the sequence of gestures needed to communicate different words and phrases. Auslan the Australian Sign Language is comprised of numbers, finger spelling for words used in common practice and a medical dictionary. There are 7415 words listed in Auslan website. This research article tries to implement recognition of numerals using a computer using the static hand gesture recognition system developed for consumer electronics control at the University of Wollongong in Australia. The experimental results indicate that the numbers, zero to nine can be accurately recognized with occasional errors in few gestures. The system can be further enhanced to include larger numerals using a dynamic gesture recognition system. © 2013 Springer-Verlag.