【摘要】 语音情感识别是从语音信号中提取一些有效的声学特征,然后利用智能计算或者识别的方法对话者的情感状态进行识别。介绍了国内外在该领域中关于语音情感数据库、特征提取、识别方法的研究现状。基于对该领域现状的了解,发现特征提取对识别率有着非常大的影响。录制了1050句语音,每句语音提取了30个特征,从而形成了一个1050×30的数据库。提出了用粗糙集理论中的信息一致性对数据库中的30个特征进行化简,最后得到了12个特征。用神经网络中的BP网络对话者的情感状态进行识别,最高识别率达到了84%。从实验结果发现不同的情感用不同的方法识别结果更好。更多还原

【Abstract】 Speech emotion recognition is about extracting effect acoustic features from speech signals and recognizing emotion state of human by using of intelligent computation.The domestic related research of emotion speech database,features extraction and recognition ways are studied.Learning from these related researches,the features extraction was found to have important affections on the speech emotion recognition.1050 sentences was recorded and 30 features extracted form every sentence and then formed to a database of 1050×30.The information consistence of rough set is applied to simplify 30 features of database to 12 features.Then artificial neural network is used to recognize emotion state of 525 sentences,it attains to the highest recognition rate of 84%.The results shows that using different ways to recognize different emotion has better effects.更多还原