Visual Interpretation of ASL Finger Spelling using Hough Transform and Support Vector Machine

Abstract: This paper proposes a vision based automated system for interpretation of American Sign Language (ASL) finger spelling using image processing. A graphical user interface is developed for the hard of hearing community to interact with normal hearing persons through hand gestures in a natural way thereby eliminating the need for an interpreter. This system does not require the user to wear any data gloves or special hardware for recognition of gestures and it provides human-machine interaction only through bare hand. In this research, Hough Transform is employed for feature extraction and gesture modeling and classification is performed through Support Vector Machine (SVM). The finger spelling of alphabets is carried out to form different isolated words. The proposed system is tested on the benchmark Triesch ASL dataset for recognition of individual alphabets and the test results show that this system achieves the overall recognition accuracy of 93.88%.

Keywords: Computer Vision, Hough Transform, Machine Intelligence, Pattern Recognition, Sign Language Recognition, Support Vector Machine.