GoogLeNet Application towards Gesture Recognition for ASL Character Identification

Gesture recognition plays a vital role in multiple applications; the applications may be related to simple aspects such as playing video games to complex tasks such as related to communication using sign symbols. In this paper, we propose a robust hand gesture recognition system using skin detection technique. Gesture recognition plays a key role for different applications. In this regard, we attempt to take out the type features of the fingers, for instance their orientation, their status in terms of raised or folded fingers and their respective image position, which require the processing of images as a matter of principle. In order to recognize the terms used for contact the program uses a complex approach to distinguish different hand movements. The dynamic sequence of the images that function as a contribution to the hand gesture recognition system is captures using the camera. GoogLeNet is used for hand gesture recognition training and detection. In order to train GoogLeNet, we use transfer learning. We also used American Sign Language (ASL) to map the sign to a particular alphabet.