With the continuous enhancement of China's comprehensive national strength and the rising sense of well-being of our citizens, reasonable concerns about the hearing impaired in China should also be considered. Yolov5, as the latest work of Yolo series, has shocked the world with its readable code, super high image processing efficiency and simple structure since it came out in 2020. The Jetson TX2 equipped with embedded GPU has the ability to realize more complex and deep neural networks. This paper uses Jetson TX2 and yolov5 algorithm as the core, uses high-definition camera to capture sign language information in real time, and then carries out target extraction and gesture recognition on video frames. The image recognition rate reaches 0.05 seconds per frame, and the average gesture detection accuracy reaches 98%. The experimental results show that the sign language recognition system based on Jetson TX2 and Yolov5 has a high recognition rate and accuracy, which is sufficient to meet the basic sign language communication of hearing impaired people in daily life.