摘要
提出了从语音信号中识别情感特征的方法 .从 5名说话者中搜集了带有欢快、愤怒、惊奇和悲伤的情感语句共 30 0句 .从这些语音资料中提取了 1 0个情感特征 .提出了 3种基于主元素分析的语音信号中的情感识别方法 .使用这些方法获得了基本上接近于人的正常表现的识别效果 .
In this paper, some methods are proposed to discriminate utterances from the speech signal. A corpus containing emotional speech of happiness, anger, surprise and sorrow with over 300 utterances from five speakers is recorded. Ten emotional features are extracted from the speech signal. Three emotion recognition methods are introduced based on principal component analysis. Using these methods, recognition performance is obtained, which is close to human performance on the task.
出处
《软件学报》
EI
CSCD
北大核心
2001年第7期1050-1055,共6页
Journal of Software
基金
国家自然科学基金资助项目 (698710 0 9)
关键词
情感识别
语音信号
多媒体
计算机
emotional recognition
speech signal
prosodic feature
principal component analysis
