Improving remote music teaching quality assessment using integrated wireless sensor networks and AGGNN-GA framework

Presently, there is a growing need to develop better ways to measure student engagement and the quality of teaching in virtual classrooms with remote learning becoming an integral part of education. Traditional methods often miss out on capturing the full range of student interactions and behaviors. To solve this, we present a new model that combines video analysis with real-time physical data collected through Wireless Sensor Networks (WSNs). The proposed scheme introduces a comprehensive, multi-modal architecture that synergizes video and sensor data through a granular attention mechanism. This enables fine-grained analysis of both facial expressions and physical behaviors, capturing the full spectrum of student engagement. At the core of the system lies an Attention-based Graph Neural Network (AGGNN), which intelligently fuses spatial–temporal features extracted from video streams with real-time physiological and motion data collected via Wireless Sensor Networks (WSNs). To further enhance adaptability and robustness, a Genetic Algorithm (GA) is embedded into the pipeline to optimize hyperparameters dynamically based on diverse educational contexts. Unlike conventional models, this proposed framework is adaptive, scalable, and continuously learns from real-time data streams by providing educators with actionable insights to personalize instruction and enhance the quality of virtual learning environments. We have performed several experiments and the results of these show that the suggested method can attain an average recognition accuracy of up to 93.6% for classroom behaviors by efficiently identifying several classroom actions.