Real Time American Sign Language Recognition Using Yolov6 Model

This research presents a real-time American Sign Language (ASL) recognition system using the YOLOv6 model. American Sign Language is the principal form of communication for the deaf and hard of hearing community, and automating its recognition holds significant societal implications. YOLOv6, known for its speed and accuracy in object detection, has been adapted for ASL gesture recognition. However, this model has been little used for this application. Our approach involves collecting a dataset of ASL gestures between numbers and letters, training the YOLOv6 model on this dataset, and fine-tuning it for real-time inference. The proposed system provides exceptional accuracy of 98.7% and speed, with an average inference time of 3.83 ms, making it suitable for applications such as real-time translation. Experimental results illustrate the model’s capability to recognize ASL signs accurately and in real-time, offering promising prospects for improving communication and accessibility for the community and society in general.