Research on Multimodal Artificial Intelligence Teaching Assistant for Intelligent Teaching in Higher Education Courses

We present a multimodal AI teaching assistant system designed for higher education teaching. By processing speech, images, and text data, this system monitors and supports classroom teaching activities. Our approach incorporates deep learning-based multimodal fusion with attention mechanisms and adversarial learning to handle noisy inputs effectively. In our tests, the system showed 92.5% accuracy in analyzing classroom engagement, surpassing traditional unimodal approaches by 15%. The system responds within 0.8 seconds on average, with 85% of users reporting satisfaction. The system also offers learning path suggestions and tutoring support, enhancing the quality of higher education teaching. Our findings demonstrate both the practical benefits of multimodal AI in education and suggest future directions for educational technology development.