Speech Recognition Method Based on Deep Learning and Its Application

Abstract

To generate better evaluation and feedback in speech recognition for language learning, this paper proposes a computer-aided training scheme. We first analyze the speech processing flow of the foreign language training and learning system based on speech recognition technology. Then HMM based speech recognition technology for feature parameter extraction is adopted for codebook generation and template training and the improved Viterbi model is used to reduce the amount of Gauss computation. Finally, the expert database is used to correct phonemes, illustrated by oral English training in real environment. With the real scene data of large-scale oral English tests, the proposed method improves the recognition accuracy by 15%, which can provide learners timely, accurate and objective evaluation and feedback guidance.