Hand Gesture Recognition using Gradient based Key Frame Extraction

Sign language is a mainstream method of communication by people suffering from hearing and speech impairment and other disabilities. Around 360 million people globally are suffering from disabled hearing loss. Minimizing the communication gap between differently-abled and normal people becomes a necessity to ensure effective communication among all. In this paper, a method for developing a system which recognizes cue symbols of sign languages to act as a communication medium between healthy and differently-abled people has been proposed. Two standard sign languages are considered for reference, as, cue symbols vary for each region. The system includes a hand gesture recognition system, where gestures are given as input and the corresponding output is in the form of text as well as speech. We have used gradient value for splitting two or more gestures from a continuous video sequence and extracting key frames. Features of pre-processed key frames are extracted using OH (Orientation Histogram) and dimensions of these features are reduced using PCA (Principal Component Analysis). SVM (Support Vector machine) classifier is used for classification of the isolated pre-processed gestures. After the classification process, corresponding connotation of gestures given as input is obtained