Hand gesture recognition using artifical neural networks

ABSTRACT HAND GESTURE RECOGNITION USING ARTIFICIAL NEURAL NETWORKS In order for humans to interact with computers, a fast and easy way is to use hand gestures. Although using hand gestures in computer interaction was very cumbersome and needed special glove and computer hardware in the past, nowadays in any personal computer, simple cameras are available and there is enough processor power to do the expensive computations done in the past. With this feature, not only controlling computer with hand is possible but also some simple interpreters for sign language recognition can be made. In this thesis a hand gesture recognition system, using an inexpensive camera and a personal computer is proposed. The system uses visual image as input and computes the geometric features, such as invariant moments and signature, of the extracted hand image for classification in a Multi-layer Perceptron Artificial Neural Network. The gestures used in the system are American Sign Language Manual Alphabet Gestures and Turkish Sign Language Manual Alphabet Gestures.