Intelligent Error Correction of Oral English Pronunciation Based on Speech Recognition Technology

Intelligent Error Correction of Oral English Pronunciation Based on Speech Recognition Technology

연구 분야: Safety

학회: 2020 13th International Conference on Intelligent Computation Technology and Automation (ICICTA)

According to the basic structure model of speech recognition system, this paper introduces the basic technology of digital signal processing and pattern recognition involved in preprocessing, feature parameter extraction and model matching. Using feature parameter extraction, forced association and dynamic time warping, speech scoring based on feature comparison and HMM is studied respectively. At the same time, Viterbi alignment algorithm, process pruning and model storage optimization are also improved, which makes the speed of speech recognition in the platform faster and it provides users better response. Finally, the performance and accuracy of the system are tested compared with the scoring algorithm based on continuous hidden Markov model. The experiment results show that the recognition speed is significantly improved. The model can basically reflect the learners' English pronunciation level, correct most of the learners' errors, and it is helpful to the training of English pronunciation level.