<p>To tackle the challenges inherent in traditional learning behavior recognition and enhance teaching quality, a classroom situational learning behavior recognition Model is introduced. This Model leverages the advancements of You Only Look Once version 5 and incorporates optimizations based on ShuffleNet V2. The findings of this research hold substantial promise for elevating teaching quality. By employing this Model, intelligent identification and analysis of student behaviors within the classroom environment are enabled, thereby furnishing teachers with supplementary insights. This capability empowers educators to gain a deeper understanding of students learning states, ultimately leading to improved teaching outcomes. The proposed Model undergoes rigorous testing, with experimental results affirming its exceptional performance. Specifically, the learning behavior detection model achieves an impressive accuracy rate of 99.2% and a recall rate of 99.4%. Additionally, a learning behavior classification network, optimized on the foundation of ShuffleNet V2, is developed. In this network, the Rectified Linear Unit activation function is substituted with HardSwish, resulting in the classification Model attaining an accuracy of 98.1%. Upon integrating these two components, the model is capable of accurately identifying five distinct student movements: prone position, raising hand, turning, standing, and the general state. This integrated approach yields commendable recognition results. When compared to alternative object detection algorithms, this Model stands out by achieving the highest accuracy and precision, with scores of 98.1% and 99.2%, respectively, while simultaneously maintaining a compact memory footprint. Positive experimental outcomes underscore the superior performance of the proposed model in learning behavior recognition within classroom settings.</p>