摘要 在日常生活中，情感是人们在学习、生活和人际交往中必不可少的。不同的 情感通常会对人们的情绪产生不同的影响从而影响人们对学习生活的态度。所以 在现代高速率的生活中怎样识别和把握人们的不同情绪成了社会上很重要的研 究探索方向。首先，要想对这个课题进行研究，我们可以从一些现实的事例中入 手，从而找到突破点。我们先进行一个实验，这个实验是利用视频的刺激效果来 进行情感的诱发，在实验的同时用现代化科技脑电信号采集装置来得到实验中实 验对象产生的脑电信号。在实验之前要将初始的脑电数据进行过滤，然后得到原 始脑电图特征带，也就是对δ电极，θ电极，α电极，β电极和γ电极 5 个电极 处于五波段时候其频带能量的对数。但是通常在实验时会有其他无关的情绪会对 实验产生干扰，实验时我们采取线性动力系统来对脑电波信号进行平滑处理，从 而可以获得所需要的脑电特征结果。利用一些技术手段，我们找到了和情感有紧 密联系的前 50 个共性特点，这些共性特点不会受实验的影响，用这些特征我们 可以找到每个特征对应的电极位置，从而得知情感在脑区的大概分布。利用现在 分析方法中的主成分分析来研究之前我们得到的特征的其他维度，采用向量机对 这些维度进行分类研究，结果在 88。5%左右。得到数据之后，再创建学习模型， 对于本课题的研究，我们需要创建的模型是流形学习，完成实验中曲线情感的的 探索研究。72734

该论文有图 10 幅，表 5 个，参考文献 27 篇。

毕业论文关键词：情感脑电波FFT向量机流行学习模型

Emotion Recognition Research Based on Eeg Signals

ABSTRACT In daily life, the emotion is people in the study, life and interpersonal communication indispensable。 Different emotions often different  influence  on people's emotions that affect people's attitude to learning life。 So in modern  high speed life how to recognize and grasp the people of different emotions became a very important research direction on the society。 First of all, if you want to study of this subject, we can start from some real cases, so as to find a breakthrough。 We first conducted an experiment, the experiment is to use the video effect of stimulation for emotional trigger, with modern technology at the same time in the experiment of brain electrical signal acquisition device for subjects in the experiments of eeg signals。 Prior to the initial eeg data filtering, and then get the original eeg features, which is on the delta electrode, theta electrodes, alpha electrodes, beta and gamma electrode  electrode 5 electrode in five  logarithm of wave when the  frequency band energy。 But normally, in the experiment, has other irrelevant emotion can produce interfe rence to experiment, experiment, we adopt linear dynamic system to smooth brain signals processing, which can obtain the required electrical characteristics。 Using some technical method, we found and emotion are closely contact the top 50 common characteristics of these general characteristics not affected by the experiment, using the features we can find each corresponding electrode position, thus learned that emotional about distribution in the brain。 Using analysis method of principal component analysis to study now before we get the characteristics of the other dimensions, using vector machine (SVM) to classify these dimensions, the result is about 88。5%。 Get data, and then create a learning model, for this topic research, we need to create a model is the manifold learning, to complete the experiment curve emotional  exploration  research。

Key words:EmotionalBrain  wavesFFTSVMManifold   learning  model