AI-based English teaching cross-cultural fusion mechanism

Due to the differences between eastern and western cultures, many students often face difficulties in language learning caused by cultural differences in learning English. Adaptability in cross-cultural communication and fluency in language learning are the highest realm of learning English and the highest pursuit for every student's learning. To push the dialogue contents for students, this paper proposes a recognition model of multi-input feature extraction based on convolutional neural networks (CNN) for English emotion recognition. The model adopts an end-to-end training method for learning, which can simplify the workflow and improve the overall recognition accuracy. At first, a speech is transformed into a form more suitable for model learning: spectrogram and Mel spectrogram. Then, it is input into the model for automatic feature extraction. In the model design, the information of spectrogram in time-domain and frequency-domain is considered at the same time. Finally, the high-level features of the image are fused, which greatly improves the effect of speech emotion recognition. Experimental results prove that the proposed speech emotion recognition fusion model has high recognition accuracy and quality of experience.