基于脑电的情感识别

关键词：脑电，情感识别，主成分分析，功率谱密度，关键频带
基于脑电的情感识别
EEG-BASED EMOTION RECOGNITION
ABSTRACT
Human emotional states significantly affect the human cognition and human behaviour. Brain mechanisms of emotion and, in particular the relationship between emotion and cognition and its neural mechanisms, has gradually become a research hotshot of neuro-science research. Emotion recognition based on computer technology is the key technology for the high level human-computer interaction. It is essential for human-computer interaction, human - computer interface and intelligent computers. This research project focuses on the variation of EEG at different emotional states. We use music emotionally different as emotional stimuli, to induce the subjects to produce either exciting or calm emotions. Using EEG power spectral density (PSD) as characteristics, principal component analysis(PCA) algorithm to reduce dimension, we use support vector machine(SVM) model for EEG classification. We found DASM 12(differential asymmetry of 12 electrode pairs) features and RASM 12(rational asymmetry of 12 electrode pairs) makes sense for emotional classification; principal component analysis (PCA) algorithm can greatly improved operational efficiency; at the same time, I found that gamma-band signal plays an important role in emotion recognition, beta-band and delta-band signal are also useful for emotion recognition, whereas theta-band and alpha-band don't perform well in emotion recognition.
Key words: EEG, emotion recognition, PCA, PSD, key frequency band
基于脑电的情感识别
目 录
第一章 绪论 ---------------------------------------------------------------------------------------------------- 1 1.1 研究背景及意义 ------------------------------------------------------------------------------------ 1 1.2 国内外研究现状 ------------------------------------------------------------------------------------ 1 1.3 工作介绍 --------------------------------------------------------------------------------------------- 1 1.4 本文结构 --------------------------------------------------------------------------------------------- 2 1.5 本章小结 --------------------------------------------------------------------------------------------- 2 第二章 生理背景 ---------------------------------------------------------------------------------------------- 3 2.1 大脑结构及功能 ------------------------------------------------------------------------------------ 3 2.1.1 大脑主要结构 -------------------------------------------------------------------------------- 3 2.1.2 大脑皮质 -------------------------------------------------------------------------------------- 3 2.2 脑电产生机制 --------------------------------------------------------------------------------------- 4 2.2.1 皮层椎体神经元突触后电位学说[4] ----------------------------------------------------- 4 2.2.2 丘脑与大脑皮层的网络学说[5] ----------------------------------------------------------- 4 2.2.3 局部规模和大规模同步学说[6] ----------------------------------------------------------- 4 2.3 脑电图 ------------------------------------------------------------------------------------------------ 4 2.4 脑电频率特性 --------------------------------------------------------------------------------------- 5 2.5 本章小结 --------------------------------------------------------------------------------------------- 6 第三章 情感研究 ---------------------------------------------------------------------------------------------- 7 3.1 情感及情感模型 ------------------------------------------------------------------------------------ 7 3.1.1 情感 ------------------------------------------------------------------------------------------- 7 3.1.2 情感模型 ------------------------------------------------------------------------------------- 7 3.2 诱发情感的刺激材料 ------------------------------------------------------------------------------ 7 3.2.1 视觉刺激材料 ------------------------------------------------------------------------------- 8 3.2.2 听觉刺激材料 ------------------------------------------------------------------------------- 8 3.2.3 嗅觉刺激材料 ------------------------------------------------------------------------------- 8 3.2.4 多媒体刺激材料 ---------------------------------------------------------------------------- 8 3.2.5 实验性场景 ---------------------------------------------------------------------------------- 8 3.3 情感识别方式 --------------------------------------------------------------------------------------- 8 3.3.1 传统识别方法 ------------------------------------------------------------------------------- 8 3.3.1.1 人脸情感识别 ----------------------------------------------------------------------- 8 3.3.1.2 语音声调情感识别----------------------------------------------------------------- 9 3.3.1.3 语言文字情感识别----------------------------------------------------------------- 9 3.3.2 利用生理信号识别情感 ------------------------------------------------------------------- 9 3.4 本章小结 --------------------------------------------------------------------------------------------- 9 第四章 实验 -------------------------------------------------------------------------------------------------- 10 4.1 实验目的 ------------------------------------------------------------------------------------------- 10 4.2 刺激材料 ------------------------------------------------------------------------------------------- 10 4.3 被试情况 ------------------------------------------------------------------------------------------- 10
基于脑电的情感识别
基于脑电的情感识别
摘要
Байду номын сангаас
人类的情感状态在很大程度上影响了人类的认知和行为。情感及其脑机制的研究，特 别是情感与认知的关系及其潜在的神经基础，已经逐渐成为神经科学的研究热点研究领域。 利用计算机技术进行情感识别是实现高级人机交互的关键技术，对于实现人机交互、人— 计算机接口以及智能计算机等有重要意义。本课题的研究重点为人类在不同情感状态下脑 电信号的变化规律。我们使用带有不同感情色彩的纯音乐作为刺激材料，诱发被试人员产 生兴奋和平静两种情感。利用脑电信号的功率谱密度特征，在使用主成分分析算法降维后， 用支持向量机模型对脑电信号进行分类。通过研究，我们发现 DASM 12 特征（12 对对称的 电极平均功率之差） 以及 RASM12（12 对对称的电极平均功率之商）对于人脑情感分类很 有意义；主成分分析(PCA)算法可以在不降低分类准确率的情况下大大提高运算效率；与此 同时，我还发现，对于情感识别来说，Gamma 波段识别效果最好，Delta、Beta 也有重要作 用，Theta、Alpha 波段识别效果较差。