EEG-Based Attention Monitoring with CNN-LSTM for Detecting Cognitive Engagement in STEM Smart Classrooms
Abstract
EEG-Based Attention Monitoring using CNNLSTM proposes an automative framework to recognize cognitive engagement in STEM smart classrooms to facilitate a better learning process. Current techniques, e.g. behavioral observation, questionnaires and conventional machine learning on EEG signals, typically suffer subjectivity, poor temporal resolution and low scalability. The proposed approach presents a hybrid CNN-LSTM network that can extract the spatial features with the help of convolutional network layers and temporal dependencies with the help of LSTM network layers to give a strong attention classification. The system preprocesses EEG signals, parts them and assesses real-time levels of engagement. The model was implemented on the Python platform with TensorFlow and Keras that demonstrated high accuracy in the classification of engaged and non-engaged states. It is an effective surveillance device suitable to educators as it allows teaching adjustments to the students to know when they are not engaged to implement individual intervention. The researchers prove that EEG-based hybrid deep learning approaches are feasible and effective to real-time cognitive assessment in the classroom.