Teaching English pronunciation in an English as a Second Language (ESL) context involves tailored strategies to help learners accurately produce sounds, intonation, and rhythm.
This study presents an innovative method utilizing advanced technology and algorithms to enhance pronunciation accuracy, fluency, and completeness. The approach employs multi-sensor detection methods for precise data collection, preprocessing techniques such as pre-emphasis, normalization, framing, windowing, and endpoint detection to ensure high-quality speech signals. Feature extraction focuses on key attributes of pronunciation, which are then fused through a feedback neural network for comprehensive evaluation. The experiment involved 100 college students, including 50 male and 50 female students, to test their English pronunciation.
Empirical results demonstrate significant improvements over existing methods. The proposed method achieved a teaching evaluation accuracy of 99.3%, compared to 68.9% and 77.8% for other referenced methods. Additionally, students showed higher levels of fluency, with most achieving a level of 4 or above, whereas traditional methods resulted in lower fluency levels. Spectral feature analysis indicated that the amplitude of speech signals obtained using the proposed method closely matched the original signals, unlike the discrepancies found in previous methods.
These findings highlight the effectiveness of the proposed method, showcasing its ability to improve pronunciation accuracy and fluency. The integration of multi-sensor detection and neural network evaluation provides precise results, outperforming traditional approaches.