A Machine Learning Based Approach for Hand Gesture Recognition using Distinctive Feature Extraction

In a world of almost 7 billion people more than 500 million suffer from some physical, sensory or mental disability. Their lives are often impeded by such deformities which bars them from full participation in society and the enjoyment of equal rights and opportunities. Sign language is common for the deaf and the dumb. Sign language is an efficient alternative to talking, where the former is replaced by hand gestures. Hand gestures are combination of hand shapes, orientations and movement of the hands, alignments of the fingers and positioning of the palm which are used to express fluidly a conveyer's thoughts. Signs are used to communicate words and sentences to audience. The objective of this paper is to optimize an algorithm for recognition of hand gestures with reasonable accuracy, where the input to the pattern recognition system will be given from the hand. Possible reference models are already available such as ASL or American Sign Language. Image is collected from a webcam followed by preprocessing. Further segmentation of the figure is done through polygon approximation and approximate convex decomposition. Feature extraction is done by recording the unique feature among the various convex segments of the hand. The resultant singularities are then used as extracted feature vectors. This involves training with the obtained features which are approximately unique for different hand gestures. Thus, we will be able to identify sign languages and successively make disabled individuals socially acceptable.