Enhancing English Oral Teaching through Pyramidal Convolution Shuffle Attention Neural Network and Sea-Horse Optimizer using Virtual Reality Technology

Abstract

The ultimate objective of teaching English to students is to help them become self-sufficient language learners and users, proficient in efficient language learning techniques, and capable of transmitting information in English. As a result, good English language instruction requires language communication training for both students and teachers, as well as between students. Compared to classroom instruction, English learning could easily facilitate English learning and provides a comfortable environment by reducing the drawback of the conventional classroom, which could lead to lower ratings for mental strain, absence of communication, and fear of making mistakes. To avoid these challenges, the pyramidal convolution shuffle attention Neural Network with sea-horse optimizer is proposed for classifying pronunciation, speaking proficiency, fluency, and intonation, of the English oral teaching. Initially, the data’s are gathered via the dataset of oral English teaching in virtual reality dataset. Afterward, the data’s are fed to pre-processing. In pre-processing segment; it removes the noise and enhances the input images utilizing federated neural collaborative filtering. The pre-processing output is fed to Feature extraction segment. Here, four statistical features such as kurtosis, mean, skewness, and standard deviation are extracted based on Adaptive and concise empirical wavelet transforms. After that, the extracted features are given to the pyramidal convolution shuffle attention neural network optimized with sea-horse optimizer algorithm for effectively classify the pronunciation, speaking proficiency, fluency, and intonation. The proposed EOT-VRT-PCSANN-SHO approach is implemented in MATLAB. The performance of the proposed EOT-VRT-PCSANN-SHO approach attains 99%, 98%, 97.5%, and 97%, as high accuracy, 98%, 98.5%, 95%, and 99% in F1 score, and 98.7%, 98%, 99%, and 97.5%, in precision, are high, when compared with existing methods.