In online education, students' real-time movements can accurately reflect their current learning state. In the case that it does not affect the study attention and ensure the security of personal privacy information, accurate identification of learning actions is a key factor in monitoring the quality of online education. This paper proposes a network learning action recognition system LD-identify based on passive RFID. LD-identify only uses radio frequency signals to complete student movement identification, so the identification system can protect personal privacy information well and prevent a series of problems such as expensive equipment. By extracting effective features of phase and signal strength, LD-Identify can achieve good performance of recognition accuracy with deep learning algorithm. The experiment shows that only two radio frequency tags sticked on the back of the hat can well identify three movements: looked up and down, shake your head around, and leaning backward. In order to further verify the performance of the system, the accuracy of six volunteers' action recognition in different scenes is investigated. The experimental results show that LD-Identify can well identify three actions of all users in different scenarios, the convolutional neural network is used to construct a classification model to recognize actions and achieve good recognition rate, and the recognition accuracy reaches more than 95.5%.