Method of intonation conversion for facilitating English communication between native and non-native speakers

Abstract

With the growth of the Internet and globalization, hardships in English speech communication became more serious for non-native speakers. It would be desirable to develop a speech communication device which can convert English speech into a more intelligible one for listeners. In this study, English speech was hypothesized to be more intelligible to non-native listeners by being modified into the listeners' native language speaking style. First, the characteristics of English between native and Japanese speeches were statistically analyzed. The analysis determined the modification model marked modification of intonation of native speakers' speech so as to include Japanese characteristics. The hypothesis was realized by speech modification generated by a speech analysis, modification and synthesis system, STRAIGHT. Then, the modified speeches were evaluated on the basis of listening experiments involving Japanese listeners. The experiments consisted of dictation tests and preference tests, and resulted in an increase in the scores. The results, therefore, verified the hypothesis. This also revealed the potential feasibility of the speech conversion method which facilitates spoken English communication.