The precise recognition of entire classroom meta-actions is a crucial challenge for the tailored adaptive interpretation of student behavior, given the intricacy of these actions. This paper proposes a Dynamic Position Embedding-based Model for Student Classroom Complete Meta-Action Recognition (DPE-SAR) based on the Video Swin Transformer. The model utilizes a dynamic positional embedding technique to perform conditional positional encoding. Additionally, it incorporates a deep convolutional network to improve the parsing ability of the spatial structure of meta-actions. The full attention mechanism of ViT3D is used to extract the potential spatial features of actions and capture the global spatial–temporal information of meta-actions. The proposed model exhibits exceptional performance compared to baseline models in action recognition as observed in evaluations on public datasets and smart classroom meta-action recognition datasets. The experimental results confirm the superiority of the model in meta-action recognition.

The individual adaptive behavioral interpretation of students’ learning behaviors is a vital link for instructional process interventions. Accurately recognizing learning behaviors and conducting a complete judgment of classroom meta-action sequences are essential for the individual adaptive behavioral interpretation of students’ learning behaviors. This paper proposes a learning behavior classification model based on classroom meta-action sequences (ConvTran-Fibo-CA-Enhanced). The model employs the Fibonacci sequence for location encoding to augment the positional attributes of classroom meta-action sequences. It also integrates Channel Attention and Data Augmentation techniques to improve the model’s ability to comprehend these sequences, thereby increasing the accuracy of learning behavior classification and verifying the completeness of classroom meta-action sequences. Experimental results show that the proposed model outperforms baseline models on human activity recognition public datasets and learning behavior classification and meta-action sequences completeness judgment datasets in smart classroom scenarios.

Purpose: Learner posture is a key non-verbal behavioral cue in smart education, reflecting engagement, cognitive state, emotional response and collaboration quality. While posture recognition has been widely studied in computer vision, its systematic integration into education research remains limited. This paper aims to provide the first comprehensive review of learner posture recognition from both technological and educational perspectives, thereby establishing a conceptual and methodological foundation for future intelligent teaching systems. Design/methodology/approach: Adopting a systematic literature review, we analyze key aspects of posture recognition – including representation models, keypoint detection, data collection, multimodal fusion and deployment strategies – while explicitly situating them within situational cognition theory. Unlike prior studies that emphasize technical performance alone, this review examines how posture recognition technologies can inform educational interpretation and intervention. Findings: Posture recognition supports real-time teaching feedback, personalized assistance and collaborative analysis. However, comparative evaluation with existing automated interaction systems and situationally aware design frameworks remains scarce, limiting the ability to assess substantive breakthroughs. By synthesizing current advances and identifying these gaps, the review clarifies both the state of the art and the pathways toward robust, theory-informed applications. Research limitations/implications: Rather than proposing yet another incremental model, this paper contributes originality through (1) offering the first taxonomy of posture recognition tailored to smart education, (2) bridging technical modeling with educational theory and (3) identifying critical gaps in benchmarking and comparative evaluation. These contributions provide an essential reference for advancing posture perception as an interpretable and educationally meaningful component of smart learning environments. Originality/value: This paper systematically integrates the research context and key technologies of learner posture recognition in smart education for the first time, proposes future development directions and has important reference value for promoting the integration and innovation of posture perception and intelligent teaching.

A student’s active thinking state directly affects their learning experience in the classroom. To help teachers understand students’ active thinking states in real-time, this study aims to construct a model which characterizes their active thinking states. The main research objectives are as follows: (1) to achieve accurate classification of the cognitive levels of in-class exercises; (2) to effectively quantify the active thinking state of students through analyzing the correlation between student cognitive levels and exercise cognitive levels. The research methods used in this study to achieve these objectives are as follows: First, LSTM and Chinese-RoBERTa-wwm models are integrated to extract sequential and semantic information from plain text while TBCC is used to extract the semantic features of code text, allowing for comprehensive determination of the cognitive level of exercises. Second, a cognitive diagnosis model—namely, the QRCDM—is adopted to evaluate students’ real-time cognitive levels with respect to knowledge points. Finally, the cognitive levels of exercises and students are input into a self-attention mechanism network, their correlation is analyzed, and the thinking activity state is generated as a state representation. The proposed text classification model outperforms baseline models regarding ACC, micro-F1, and macro-F1 scores on two sets of exercise datasets in Chinese containing mixed code texts, with the highest ACC, micro-F1, and macro-F1 values reaching 0.7004, 0.6941, and 0.6912, respectively. This proves the proposed model’s effectiveness in classifying the cognitive level of exercises. The accuracy of the thinking activity state characterization model reaches 61.54%. In particular, this is higher than the random baseline, thus verifying the model’s feasibility.