作者: Aijun Li , Haibo Wang
DOI:
关键词: Duration (music) 、 Natural language processing 、 Chinese speech synthesis 、 Artificial intelligence 、 Speech technology 、 Perception 、 Standard Chinese 、 Expression (mathematics) 、 Tone (linguistics) 、 Computer science 、 Speech recognition
摘要: Previous analyses on friendly speech were made using dialogues without differentiating their linguistic functions. This paper reports an analysis declarative and interrogative sentences respectively. Pitch duration of prosodic words statistically analyzed compared concerning factors positions stresses. Based the acoustic investigation speech, tonal pitch distributions adjusted in synthesis synthesized stimuli subjected to perception test. It was found that: (1) Friendliness could be achieved via adjusting perceptually distinctive parameters; (2) Tonal is most important cue for better expression friendliness; (3) Only no use expressive speech; (4) Interrogative got higher perceptual results than sentences; (5) A high boundary tone sentence usually used by speakers express speech.