Classifying gestures by using a self-organizing neural network

One of the ways of improving the information technology systems is by allowing the computer user to communicate in a more natural way than through the use of the keyboard. One of the approaches to developing such novel input devices is the recognition of gestures. In this sense, the main goal of this paper is to develop an adaptive interface to make the communication easier for people with psychomotive disabilities. The developed prototype uses gestural input. Animation software has been used for obtaining a wide range of gestures. To both use these gestures as input patterns and to associate them with their appropriate output we use the learning and generalisation properties of self-organising neural networks. This type of neural network learns to recognise the features from the gestures obtained by the animation software; in addition it can generalise from this information to classify other inaccurate or ill-formed gestures.