In modern society, the development of the mobile Internet is getting faster and faster, and people's lives have become more and more digital. Various sensors, such as cameras, are also slowly entering people's lives, so that people can use programs and algorithms through digital technology. To understand the world and change the world, thereby significantly improving the level of information digitization and intelligence. With the rise of all kinds of short video software, video is more and more widely spread on the Internet, and users can understand, analyze and study classical hand dance movements through video. This not only helps users to strengthen the cognition and research of Chinese classical hand dance, but also contributes to the popularization and dissemination of Chinese classical traditional culture and art, thereby enhancing the cultural confidence of the nation and the country. Therefore, this paper mainly combines deep learning, studies the classical hand dance action extraction and analysis method in video, and finally designs and develops a classical hand dance action extraction and analysis system with simple equipment, convenient operation and smooth operation.