Motion Behavior Recognition Algorithm and Case Analysis Based on OpenPose

As a powerful open-source human pose estimation library, OpenPose, with its unique deep learning algorithm, can quickly and accurately detect the key points of the human body in images or videos, and plays a pivotal role in the field of motion behavior recognition, becoming the preferred tool for many researchers to carry out related research and application development. This research is based on OpenPose algorithm, aiming at the existing problems of this algorithm, it has improved from the two aspects of introducing SE and multi-modal data fusion. Basketball is selected as an example, and a variety of advanced equipment and methods are used to ensure that the collected data is highly reliable and widely representative. Using the same experimental environment, the shooting action, dribbling action and passing action of the collected data are identified respectively. The results show that the improved model can better deal with the complex and changeable action scenes in basketball.