普通话的情感语音韵律分析.pdf

摘要摘要随着人机交互技术的快速发展,近年来语音信号中的情感信息处理成为语音研究的热点之一,本文主要对情感语音韵律进行分析。本研究使用声学分析结果与感知实验结果相结合的方法、对比分析法及统计分析法,在建立相应情感语音数据库的基础上,对情感语音韵律进行了较为深入的研究和探讨。通过研究,本文发现:高兴、愤怒、悲伤、惊奇和中性语音在韵律特征如语调、语句时长、语句基频最大值、语句基频最小值、重音位置时长等方面存在一些显著性差异,具体表现如下: 重音听辨结果上:愤怒语句重音出现频率最高,悲伤语句重音出现频率最低。高兴重音多位于句尾,愤怒重音多位于句中,悲伤重音多位于句中、句尾位置; 惊奇句尾重音居多。在语句基频与声调上:高兴、愤怒、悲伤、惊奇、中性调域大d,Jl页序为:愤怒>惊奇>高兴>中性>悲伤;高兴与愤怒相比基频高低值之间的过渡平缓;高兴语调以为高调阶窄调域为主;愤怒基频多分布在高频段,语调以高调阶宽调域为主;悲伤的基频整体变化不大且多分布在低频段,语调以低调阶窄调域为主,各位置处音节较多地保持了单字的调型;惊奇句尾基频上升,语调以高调阶宽调域为主。在句重音音节基频表现上:愤怒重音位置处调域拓宽幅度最大,其次为惊奇。悲伤和中性略有拓宽,高兴重音位置处调域拓宽幅度大于悲伤和中性。在时长特征上:高兴、愤怒、悲伤、惊奇、中性语速快慢为:愤怒>高兴> 中性>惊奇>悲伤;愤怒重音时长占整句话时长比例最大,悲伤重音时长占整句话时长比例最小,惊奇重音位置处音节时长增幅最大。在停顿上:高兴停顿次数最少,停顿总时长最短;悲伤停顿次数最多,停顿总时长最长。本文创新点主要有: 为检验语料质量,进行了情感类别、情感表达程度和情感焦点的听辨实验, 使语料库信息更完善、全面,便于深入研究。本文使用声学分析结果与听辨实验结果相结合的方法,对情感语音韵律特征进行分析和描述,增强了分析的理据性。关键词:情感语音听辨实验语调分析声调分析 Abstract Together with the rapid technological developments in puter interaction in recent years,emotional information processing within speech signals ed one ofthefocalpoints ics paper primarily deals withanalysing theprosody ofemotional speech. Inthisstudy,uses the method bined theresultsofacoustic analysis and perceptual experiment,parative analysis and statisticalanalysis,based on the establishment ofthecorresponding emotional speech database and more in。depth study and discussion inemotional speech prosody. We found through research:Happy,anger,sadness,surprised and neutral tone of voice characteristics,there are some significant differences intheprosody,such as duration,the maximum fundamental frequency,minimum fundamental frequency, accented tone location,duration, performance is asfollows: IIltheresultsof stress perception:Emotional stress themost frequent manger, sadneSS thelowest frequent ofemotional than happy intheend of the sentence stress;Anger stress more inthemiddle ofsentences;Sadness more stress In themiddle and endofsentence;Surprise stressmore fortheend ofthe sentence· Infundamental frequency(E0)of sentences andto