Research on multimodal interaction of deep learning-based educational robots in English listening and speaking training

With the rapid development of intelligent educational technology, the application of educational robots in English listening and speaking training has received widespread attention. The use of educational robots in the process of English teaching helps to improve students' attention, active communication ability and learning enthusiasm in English learning, and plays a very important role in optimizing the teaching environment. Existing robot interaction modes are mostly single-modal, which is difficult to fully meet the actual needs of language learning. In this study, a multimodal interaction model is designed and optimized with deep learning technology as the core, aiming to improve the interaction experience and training effect of educational robots in English listening and speaking training. By integrating speech recognition, image processing and semantic analysis technologies, a real-time and efficient interaction system is constructed, and tested and verified in real teaching scenarios. The model significantly improves the learners' language expression ability and learning interest, and at the same time provides new ideas and technical support for the further application of intelligent educational robots in the field of language teaching.