Optimization of computer-aided english pronunciation training data analysis system

In this paper, we propose an audiovisual fusion method based on the optimization of a computer-aided English pronunciation training data analysis system, which is based on the Convolutional Neural Network (CNN). An independent CNN structure is utilized to achieve independent modeling and asynchronous information transfer of audiovisual perception and to obtain descriptions of audiovisual parallel data in high-dimensional feature space, and then the long-time dependencies of the audiovisual parallel data in higher dimensions are modeled through a shared full-connection structure immediately following the CNN. Constructing auditory features to visual features characteristics of the generative model; the generative model is then used to automatically generate many visual features, which are combined with the CNN-based audiovisual fusion method to perform bimodal modeling. The experiments show that when the generated model is trained and tested in the same acoustic environment, only a small amount of audio-visual parallel data is required, and in combination with the proposed bimodal method based on visual feature generation. The bimodal method based on visual feature generation can effectively solve the problem of missing visual information in the actual usage environment. The audiovisual fusion method proposed in this paper can model the independence, asynchrony, and long-term interdependence between audiovisual parallel data, which is of great significance for the further study of the audiovisual fusion method based on deep learning. © 2021 CAD Solutions, LLC.