English Speech Recognition and Evaluation of Pronunciation Quality Using Deep Learning

Abstract

English is now one of the most important languages for economic exchange in various countries around the world, and it is also the most widely used language for cultural and information exchange. Like other countries, China likewise attaches highest significance to English learning, and people’s demand for applied learning is also increasing rapidly these days. However, there are significant differences between Chinese pronunciation and English pronunciation, and China lacks an English environment while teaching English language. Furthermore, the traditional education is limited by the place and time of classes, due to which it cannot meet people’s needs for learning English. With the fast progress of computer knowledge, the emergence of deep learning technology can better identify English pronunciation and evaluate the quality of English pronunciation. Additionally, deep learning can provide learners with precise, objective, and rapid pronunciation information. It can also assist learners in determining the differences between their pronunciation and conventional pronunciation through frequent listening and comparison, as well as correcting their pronunciation faults and increasing language learning efficacy. This study looks into the difficulty of using deep learning to evaluate the quality of English speech recognition and pronunciation. To evaluate English pronunciation quality, this paper selects intonation, speed, and rhythm, as the distinguishing indicators. The comparison between the results of manual evaluation and our evaluation clearly shows that English speech recognition and pronunciation quality model using deep learning established in this paper has much higher reliability. Among the 240 samples tested, only 32 samples differ by one grade, and the rest are similar.