الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 100 |  |  |
| | | from | 100 | | |
Abstract
This work presents the results of developing a real hand-shape classification system, several models for Static Hand Recognition such as Individual LVQ based on DCT or Projection Features and Combined Multi-Stage LVQ Classifiers has been introduced. Our Experiments were done on Thomas Moeslund’s Gesture Recognition Database [20]. It produce a chain of relatively complex pre-processing steps to remove user dependant features such as color and scale and prepare the patterns for further steps. Two different feature extraction techniques; Discrete Cosine Transform and image Projection have been used that convert data into a new form which is suitable as an input to the neural network to be trained without loss of the information represented on the original image. A series of tests were then performed to experimentally evaluate our models. During the course of our experiments we endeavored to identify particular techniques and parameter values that improved the accuracy of our system. Future Work Although many researches has been published in recognizing hand gesture, it still suffers from lack of satisfaction and being far from complete. As the work showed, our proposed system concentrates on recognizing hand-shape in a static format and still effected by rotation transformation. In the future , we wish to acquire data from multiple sensor and to keep track of hand configuration, relation between hands and direction of the hands motion in dynamic format. Also, we wish to extend the system to handle the rotation transforamation.