2019 Volume 2019 Pages 229-234
Currently, we are developing English pronunciation training system for Japanese adults. In this system, pronunciation is indicated by displaying formants, lip shapes, and tongue shapes that are effective for identifying vowels. In this work, focusing on the visualization of tongue shape and lip shape, analysis of MR image and speech signal was performed. Generally, there is a correspondence between the first formant and the opening length of the mouth. Therefore, we obtained correlation coefficients of the first formant detected from speech signal and opening of lips and jaws detected from MR image. As a result, it was found that some cases showed strong correlation or not depending on subjects. From past research, it is known that the Japanese vowel /u/ which is the lingual tongue vowel may change to the front tongue due to the protrusion of the lip. From this, it is considered that the subject who don't show strong correlation manipulates the phoneme by projecting the lips instead of opening the mouth vertically. In order to support this hypothesis, it is necessary to analyze more MR images and speech signals in the future.
