Design of intelligent speech recognition and pronunciation correction system for higher vocational English education

Aiming at the difficulties and pain points in pronunciation teaching in higher vocational English education, an intelligent speech recognition and pronunciation correction system is designed. Based on the modular architecture, the system includes five modules: speech input, speech recognition, pronunciation analysis, pronunciation correction and feedback. Through the orderly interaction and data flow of each module, students' pronunciation problems can be automatically identified, analyzed and corrected. The system adopts deep learning algorithm, combines MFCC feature extraction, CNN-RNN mixed model architecture and CTC loss function to realize high-precision speech recognition, and generates targeted correction suggestions through technologies such as dynamic time warping (DTW) and phoneme error rate determination. The test results show that the word error rate (WER) of the system is less than 8% in quiet environment and less than 15% in noisy environment, and it is outstanding in phoneme error correction, with the acceptance rate and error correction rate reaching 82.3% and 76.5% respectively. The overall performance has met the actual needs of higher vocational English education and can effectively improve students' pronunciation accuracy and autonomous learning ability.