Static Gesture Recognizer Using Hybrid Neural Network

The main objective of this paper proposes an Embedded System employing Hybrid Neural Network (NN) for an Efficient Static Recognizer. The Hybrid Neural Network consists of Active Contour Model (ACM) and Convolutional Neural Network (CNN) in which the input data of hand sign is pre-processed and segmented using ACM and that image is feed forward to CNN classifier to classify the image for free-air hand gesture recognition without the use of hand gloves or sensors to offer a more efficient way of gesture segmentation as gesture boundaries. The system is developed to detect the hand signs in American Sign Language (ASL), and its convenience will be verified through simulations and the input is converted from sign to text and vice versa. The image is described in four languages such as English, Tamil, Hindi and Telugu and there is a need in designing an efficient human-computer interface. The possibility of the system in which the hand shapes or poses of static gestures is being captured by a camera, to improve the way of conveying information’s between the deaf and the non-deaf.