Transforming Sign Language using CNN Approach based on BISINDO Dataset

The rapid growth of technology has a significant role in developing technologies that can support communication for hearing or speech impaired individuals that enables a real-time translation into visual or written language visually. These rapid growth and advancement create an opportunity to connect the gap between impaired individuals and others for the sake of fostering the society. Sign language is commonly used by hearing or speech impaired people to communicate with others. BISINDO is Indonesian Sign Language used by deaf individuals in Indonesia. Based on BISINDO proposed a convolutional neural network model to detect and recognize sign language to accommodate a communication between impaired people. CNN is a deep leaning algorithm designed specifically to handle processing and analyzing visual data in terms of images and video. It consists of various convolutional layers that extract and analyze local features and patterns to allow object recognition and image classification. In addition to that CNN is the best possible and ideal algorithm due to their capabilities of capturing information, learning patterns and extraction with variety of hand gestures while achieving them in real time. Hence CNN is used in our model, Since the main objective of this paper is to create the most accurate and efficient model from an existing dataset. We obtained the results by using CNN, by gathering necessary data, data processing, model training and testing. Based on CNN we are able to achieve an accuracy of 82.56%.