基于脑电信号样本熵的情感识别

发布时间：2018-06-01 19:11

本文选题：情感识别 + 脑电信号　；参考：《太原理工大学》2014年硕士论文

【摘要】：情感识别是当前研究的一个热点课题,属于人工智能研究领域。对人的情感和认知的研究是人工智能的高级阶段,研究人脑是如何处理各种情感状态,对于探究人脑的运作机理有着十分重要的意义。情感识别在人们日常生活中起到的作用也越来越重要,由此产生了很多针对人类情感进行研究的方法,其中脑电信号特征提取是研究人类情感的主要手段之一。基于EEG的情感识别应用非常广,论文在已有研究的基础上,着重进行了情感脑电的特征提取及识别的研究,主要工作如下： (1)提出一种基于脑电信号样本熵的情感识别方法,EEG信号经过伪迹去除和滤波处理之后,通过K-S检验筛选样本熵存在显著差异的电极,形成情感分类的特征向量,然后利用SVM-Weight算法进行分类。 (2)设计基于图片刺激材料的心理学实验范式,并采用实验室购买的BP (Brain Products)脑电信号记录仪系统在该实验范式基础上进行EEG信号的采集；然后使用BP系统提供的软件将采集到的EEG信号进行预处理,并提取出预处理后数据的p波段EEG信号；最后在提出理论的基础上针对提取出的p波段EEG信号进行正负两类情感状态的识别。 (3)筛选Deap网站提供的预处理后的情感数据,找出其中行为实验与被试标注一致的视频；然后提取出预处理后数据的p波段EEG信号；最后在提出理论的基础上针对提取出的p波段EEG信号对被试不同激活度和愉悦度的情感状态进行识别。 (4)比较与分析样本熵与其他三种特征提取方法(近似熵、LZC复杂度和Hurst指数)的情感识别正确率以及特征提取的效率,结果说明相较于其他三种特征提取方法,样本熵更适合于提取脑电特征并进行情感识别。总之,本文的结果充分表明,使用样本熵作为脑电信号特征用于情感识别具有一定的识别效果,同时证实了脑电信号的p波节律特征用于情感识别的可能性,并找出了与情感识别活动相关的脑区。期望这种方法能在BCI、智能医疗护理系统等应用领域中得到很好的应用。
[Abstract]:Emotion recognition is a hot topic in current research and belongs to the field of artificial intelligence. The study of human emotion and cognition is the advanced stage of artificial intelligence. It is of great significance to study how the human brain deals with various emotional states. Emotion recognition plays a more and more important role in people's daily life. As a result, there are many methods to study human emotion, among which EEG feature extraction is one of the main methods to study human emotion. Emotion recognition based on EEG is widely used. On the basis of existing research, this paper focuses on the research of feature extraction and recognition of emotional EEG. The main work is as follows: 1) an emotion recognition method based on EEG sample entropy is proposed. After the EEG signal is removed by artifact and filtered, K-S test is used to screen the electrode with significant difference of sample entropy to form the feature vector of emotion classification. Then SVM-Weight algorithm is used to classify. (2) designing the psychological experimental paradigm based on image stimulation material, and adopting the BP brain products EEG recorder system purchased by the laboratory to collect EEG signals on the basis of the experimental paradigm. Then the software provided by BP system is used to pre-process the collected EEG signal and extract the p-band EEG signal of the pre-processed data. Finally, on the basis of the proposed theory, the positive and negative emotional states of the extracted p-band EEG signals are recognized. (3) screening the pre-processed emotional data provided by the Deap website, finding out the video of the behavior experiment consistent with the tagging of the subjects, and then extracting the p-band EEG signal of the pre-processed data. Finally, based on the proposed theory, the emotion states of different activation and pleasure degree were identified for the extracted p-band EEG signals. (4) comparing and analyzing the sample entropy and the other three feature extraction methods (approximate entropy, LZC complexity and Hurst index), the accuracy of emotion recognition and the efficiency of feature extraction are compared. The results show that compared with the other three feature extraction methods, Sample entropy is more suitable for EEG feature extraction and emotion recognition. In a word, the results of this paper fully show that the use of sample entropy as EEG signal features has a certain effect in emotion recognition, and the possibility of using p-wave rhythm feature of EEG signal in emotion recognition is also confirmed. The brain regions associated with emotional recognition activities were also identified. It is expected that this method can be applied in BCI, intelligent medical care system and other fields.
【学位授予单位】：太原理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN911.7

【共引文献】