【作者】 王艳；

【导师】 张桂芸；

【作者基本信息】 天津师范大学 ， 教育技术学， 2012， 硕士

【摘要】 焦虑是一种负性情绪状态,表现为对未来威胁和不幸的忧虑预期,并伴随着紧张、烦躁不安或一定的身体症状。特质焦虑是焦虑的一种,是个体在焦虑易罹患性上相对稳定的个体差异。研究表明,各种情绪的变化会在不同的大脑皮层位置反映出不同的脑电信号(EEG)。EEG含有丰富的频率成分,不同的心理状态、生理状态和认知任务下某些频段的能量在头皮不同区域的分布会发生变化,因此可以提取不同频段上的能量作为分类器的特征参数实现特质焦虑者和非特质焦虑者EEG的识别与分类。本文采用独立分量分析对天津师范大学心理与行为研究院脑电实验室采集到的特质焦虑者及非特质焦虑者的EEG进行预处理,再运用小波分析对预处理数据进行特性分析和特征提取,在此基础上结合支持向量机分类器实现两类信号的识别与分类。具体内容如下：1.运用独立分量分析(ICA)对原始信号进行预处理。ICA处理的对象是相互统计独立的信号源经线性组合而产生的一组混合信号,特质焦虑EEG中的伪迹正好符合这一特点,因此ICA适用于特质焦虑EEG的预处理。2.运用小波变换对所选取的EEG进行多尺度小波分解,得到的不同尺度的频带分量,提取EEG在不同频段上的能量特征,作为分类器的输入向量。小波变换是一种多尺度信号分析方法,具有良好的时频局部化特性,很适合分析像特质焦虑EEG这样非平稳信号的瞬态特性和时变特性。3.利用支持向量机(SVM)分类器实现对两类EEG的识别和分类。SVM是根据统计学理论提出的一种机器学习方法,它在解决小样本、非线性和高维的机器学习问题中表现出了许多特有的优势,已经逐渐成为解决模式分类问题的首选工具。研究结果表明,采用本文的方法可以取得比较好的分类效果,分类准确率达86%以上。更多还原

【Abstract】 Anxiety is an emotional response, is familiar with a negative emotional state, and trait anxiety is one of Anxiety, which refers to a general personality characteristics or traits. Studies have shown that the different state of mind of emotions and changes in the cerebral cortex in different locations will reflect different brain signals. A number of frequency components are included in spontaneous EEG and the energy corresponding to different frequency bands, which is detected in different physiological states and cognitive tasks, changes with the scalp area. Thus the energy corresponding to a certain frequency sub-band can be taken as a feature parameter of the classifier to realize the Classification of Trait anxiety EEG. In this paper, independent component analysis (ICA)was used to preprocess the raw EEG, then the use of wavelet analysis was supposed for the extraction of discriminating features from trait anxiety and non-trait anxiety’s EEG. Subsequently, combined with support eigenvector machine classifier (SVM), the paper achieved the classification of two types of signal pattern. The main work done in this dissertation is as follows:1. The use of independent component analysis (ICA) to preprocess the original signal. Artifact in the EEG can be considered by the independent source generated, and each source can be considered as a mixture of Linear mixed, so ICA was applied to Trait anxiety EEG analysis and preprocess.2. Using the method of wavelet transaction to decompose the EEG recordings into various frequency bands through multi-scale decomposition, then using wavelet coefficients to extract the energy feature that will be as the classifier input vector. Multi-scale wavelet transform is a signal analysis method, which has good time-frequency localization properties. It is ideal for the analysis of non-stationary signals such as transient and time-varying characteristics.3. The use of support to the machine (SVM) to extract the feature eigenvector for training and testing, to achieve the Trait anxiety EEG recognition and classification. SVM is a machine to solve the small sample, nonlinear and high dimensional problems, which shows unique advantages in machine learning. Using SVM for classification and identification of EEG as well as brain function research is extremely important and valuable.Experimental results show that the feature parameters extracted by wavelet analysis, as the SVM input eigenvector can achieve relatively good results, classification accuracy arrives86%.更多还原