Building A Mandarin-hainanese Speech Corpus With Multi-Modal Physiological Information And Phonetic Feature Analysis

With the implementation of developing Hainan's policy,the population mobility of Hainan will further increase,and the protection and research of Hainan dialect have become very important topic.At present,the research methods of Hainan dialect are relatively primitive at home and abroad.The basic theories and techniques of experimental phonetics have not been promoted and applied in the study of Hainan dialect.So far,there is not any complete multi-modal physiological speech database of Hainan dialect.Therefore,establishing a multi-modal physiological speech database of Hainan dialect and Putonghua plays an important role for the basic research and protection of Hainan dialect.This thesis mainly focuses on Hainan Wenchang dialect.First,we design and establish a text corpus set of the Wenchang dialect and Mandarin-Hainanese,which contains 32 Hainanese sentences and 31 Chinese sentences.On this basis,we further acquire multi-modal data by means of the Terason Ultrasound and Electroglottograph system.And then,we remark manually the wave audio recorded by the system called Praat,and finally build a Wenchang dialect and Mandarin-Hainanese speech database with multi-modal physiological information.Based on the speech database above,we further clear up and analyze the multi-modal data.We firstly analyze the characteristics when Hainan Wenchang speakers pronounce the vowels of Hainan Wenchang dialect and Putonghua and collect the formant data when the different gender speakers pronounce the two kinds of vowels,and give the vowel space diagrams corresponding to the formant data.Then,we compare and investigate the vowel spaces above,and we found that the mother tongue has an effect on the Mandarin space produced by Hainanese speakers.Moreover,we compare the vowel spaces of English,Mandarin,and Hainanese produced by their respective native speakers,and it can be found that the size of the Hainanese vowel space was similar to the Mandarin and English vowel spaces,and it also shows that the number of the vowel has no obvious effect on the size of vowel space.