Application of speech emotion recognition in English teaching mode based on multi-core learning feature fusion

Abstract

Abstract Educational informatization has also had a significant impact on college English education, which has brought about a major change in the traditional English education model, and the adjusted English education has been used as an international language. Therefore, the organic combination of English education and the new education model will better promote the interaction between education and learning. In the new era of education, college English teachers should seize the opportunity to change teaching concepts, innovate teaching methods, deepen educational reforms, and make active scientific explorations of the "dual classroom" teaching model in college English teaching. This paper designs a voice emotion recognition method based on multi-core learning and multi-function feature fusion. At the same time, in order to obtain high-discrimination feature information, the motion map of speech depth is used as the feature information source, and the features of spatial multi-scale binary histogram speech and gradient histogram speech are made into three-dimensional. According to the information of the shape structure, we can extract the characteristics of speech and speech emotion in time and space, and use the Fourier transform in the time series to map the characteristics of the time series to the frequency domain. On the one hand, the feature vectors are rearranged. On the other hand, it allows us to change the form of function. In this article, we use public databases MSRGesture3D and SKIG to retrieve attribute data through a series of spatiotemporal structure attribute extraction algorithms. The results show that this method has strong anti-speech emotion ability and is very effective for speech classification of deep data. Compared with traditional feature extraction algorithms, the classification effect is better.