Sign language word translator using Neural Networks for the Aurally Impaired as a tool for communication

Abstract

Kohonen Self Organizing Maps are a type of Neural Networks which learn to recognize patterns and classify data sets in unsupervised manner. This research aimed to develop a system which converts hand gestures into Filipino words using algorithm such as Kohonen Self-Organizing Map. The system uses a webcam to capture hand images which can be processed to serve as input for Self-Organizing Map. Image processing techniques such as: color segmentation, visual-hand tracking, pre-processing, and feature extraction are used to achieve the objective. The system was developed using neural network toolbox and graphical user interface in MATLAB. The results show that the system can achieve 97.6% of recognition rate for 5 persons. It can be concluded that the system can be used by different users while generating high recognition rate.