English Numbers Recognition Based On Sign Language Using Line-Slope Features And Pso-Dbn Optimization Method

Sign Language is considered as the primary method that used by dumb and deaf people for communication. The need for a computer based system that has the ability for recognizing these signs are imperious for the dumb community. However; despite researchers have been attempted to find a solution to this problem in the previous few years but the results are still not good enough. In this paper, a system for identifying the number involved in hand gesture is introduced. Hue Saturation Value (HSV) color model is used to allocate hand region. Discriminated features are generated by finding the slope of the line that connects any two points in hand region. A combination of particle swarm optimization and deep belief network is then used to find the optimal feature subset from the generated slope features. National University of Singapore (NUS) hand posture public dataset is used for system evaluation and the achieved accuracy is 99.58% when the number of blocks is set to 7.