Combining Transformer, CNN, and LSTM Architectures: A Novel Ensemble Learning Technique That Leverages Multi-acoustic Features for Speech Emotion Recognition in Distance Education Classrooms