Optimization of Student Behavior Detection Algorithm Based on Improved SSD Algorithm

Abstract

Despite advancements in educational technology, traditional action recognition algorithms have struggled to effectively monitor student behavior in dynamic classroom settings. To address this gap, the Single Shot Detector (SSD) algorithm was optimized for educational environments. This study aimed to assess whether integrating the Mobilenet architecture with the SSD algorithm could improve the accuracy and speed of detecting student behavior in classrooms, and how these enhancements would impact the practical implementation of behavior-monitoring technologies in education. An improved SSD algorithm was developed using Mobilenet, known for its efficient data processing capabilities. A dataset of 2,500 images depicting various student behaviors was collected and enhanced through preprocessing methods to train the model. The optimized SSD model outperformed traditional algorithms in accuracy and speed, thanks to the integration of Mobilenet. Evaluation metrics such as precision, recall, and frames per second (fps) confirmed the superior performance of the Mobilenet-enhanced SSD algorithm in real-time environmental analysis. This advancement represents a significant improvement in surveillance technologies for educational settings, enabling more precise and timely assessments of student behavior. Despite the promising outcomes, the study faced limitations due to the uniformity of the dataset, which mainly consisted of controlled environment images. To improve the generalizability of the findings, it is suggested that future research should broaden the dataset to encompass a wider range of educational settings and student demographics. Additionally, it is encouraged to explore alternative advanced machine learning frameworks and conduct longitudinal studies to evaluate the influence of real-time behavior monitoring on educational outcomes.