Human Action Recognition and Analysis Methods Based on OpenPose and Deep Learning

Currently, commonly used human action recognition (HAR) methods include two categories: based on manual features and based on machine learning. However, these traditional methods often rely on handcrafted features, which require extensive domain knowledge and may not capture all the intricacies of human actions. To address this limitation, this article proposes a novel approach that combines the key technologies in OpenPose, a state-of-the-art pose estimation algorithm, with deep learning techniques. By leveraging the rich spatial and temporal information provided by OpenPose, the proposed method can capture fine-grained details of human actions with higher accuracy and efficiency. The deep learning component further enhances the recognition performance by automatically learning discriminative features from the input data. Experimental results show that the application of this method can increase the accuracy of HAR to a maximum of 95.6%. Therefore, it can be determined that the recognition method based on OpenPose and deep learning has high accuracy, precision and recall rates in HAR.