Image Processing Model for Sign Language Recognition

The act of sign language dates back to the Stone Age when there was no formally defined language in the world. It is also a means of communicating with these special people (i.e. the deaf and dumb). Research shows that a variety of sign language recognition system is already in place presently. However, there are drawbacks associated with these systems. These drawbacks range from poor quality of training databases, un-robust recognition models and lack of GUI for good visualization of sign language recognition. In this work, an enhanced image processing model for sign language recognition was developed. The system was aimed at helping in the education and interaction with the speech impaired in the learning environment. Object-Oriented Analysis and Design Methodology were adopted in this approach. The system was implemented using MATLAB. The SURF algorithm was used to perform feature extraction on the input images in order to enhance detection of interest points on the images. The results show that using our model, sign language images can be recognized at the rate of 90% accuracy as compared to the existing model with performance accuracy of 62%. This study could be beneficial to speech-impaired individuals in the educational sector, to the academic communities that support speech impaired persons, to special schools for the impaired, to organizations that cater for speech impaired persons and to the research community.