karannaik / americansignlanguagedetection Goto Github PK

American Sign Language is one of the predominantly used sign languages among all other global sign languages. It is natural language used by Deaf communities in the United States. The language is developed with the hand gestures, facial expression and motion. In this report, we present the way to detect the alphabets and digits by using Deep Convolution Neural Network. We also propose to use the pre-defined Deep Neural Network and compare the accuracy of AlexNet and VGG16 Net. We propose the implementation of the application to take the video of hand gestures and detect the alphabet and digits with an accuracy of approximately 98%. The proposed methodology is scalable and extendable to have multiple classification classes. The proposed methodology is flexible for taking input with a variety of RGB image sizes. The classification procedure is followed by pre-processing of images to extract reliable feature and remove noise if any.