3 confidence level of considering the human body action as each action type is determined based on the features extracted from the images,and an action type of a highest confidence level (higher than a preset threshold) is In the existing solution,features of all regions in the images are extracted. These features include numerons determined as the action type of the human body in the video.features unrelated to the action. Consequently,a final action recognition effect is unsatisfactory. In addition,in another existing solution,the action is recognized by directly extracting features of some regions in the images. However,an action feature of the human body may not be well reflected by directly and simply extracting the features of some regions in the images,still resnlting in relatively low action recognition accaracy.This application provides an action recoguition and pose estimation method and apparatus,to improve action According to a first aspect,an action rec