Research on Teacher Classroom Teaching Speech Emotion Recognition Based on LSTM

Speech emotion recognition technology can automatically identify the emotional states of teachers during teaching, assisting them in reflecting on their teaching behaviors and improving the effectiveness of classroom teaching. Existing research often relies on collecting facial and physiological data to recognize teachers' emotions, which presents challenges due to high costs and strong invasiveness. Addressing this issue, this study first induces specific emotional speech of teachers in real-life scenarios and establishes a teaching speech emotion corpus for teachers. Subsequently, a deep LSTM neural network model is constructed to recognize teachers' teaching emotions, with the accuracy of the model reaching 79% on the self-built dataset. Finally, the model is applied to actual classroom environments, and it is found that the model can accurately identify common emotions of teachers in the classroom, providing an objective feedback on the emotional state of teachers' teaching.