A Study on the Application of Acoustic Model for Korean Pronunciation Education

The purpose of this study is to present the method and task of using an acoustic model and a speech corpus to develop applications for Korean pronunciation education. Through preceding researches, it was confirmed that there is a lack of an application for pronunciation education using an speech recognizer, and that there is a limit to applying the pre-developed speech recognizer to pronunciation education as it is. Accordingly, a method was proposed to directly train and utilize the acoustic model, which is a part of the speech recognizer, for the purpose. In addition, with the aim of developing an application for the pronunciation discrimination practice of ‘ka’, kka, kha’, a speech corpus for training was constructed and the pronunciation recognition pattern was examined by actually training the acoustic model. With the recent development of artificial intelligence and computer technology, there was no significant difficulty in constructing a corpus and using an acoustic model. However, the recognition result did not meet the expectations, which was expected to be due to the quality problem of the speech corpus for training. As a result, it was confirmed that a high-quality speech corpus is needed to ensure the effectiveness of the acoustic model.