Real-Time Indian Sign Language Recognition System using YOLOv3 Model

Sign Language is a language which helps deaf and mute people for communication with hearing people. The aim of Indian Sign Language recognition (ISLR) is to understand the meaning of signs of speech impaired or hearing impaired person in the Indian region to interact with the society. This paper proposes for ISLR system in real-time based on the YOLOv3 Model and used in conjunction with Darknet-53 convolutional neural network. The system has been tested in real-time with 16 different signs for images and 7 signs for videos. The proposed model was labeled the sign language datasets in YOLO format. The sign language images are captured by the webcam for static sign language recognition and videos are recorded for dynamic sign language recognition. We achieved the accuracies for static and dynamic signs as 95.7% and 93.1%, respectively. The experimental results show that the proposed system can recognize both static and dynamic ISL signs effectively in real time with reduced computational time.