摘要
以面部表情视频序列为研究对象,基于一种统计学习方法-支持向量机对面部表情进行识别及强度度量。采用一种改进的特征点跟踪方法来提取面部特征形变。通过非线性降维方法-等容特征映射自动产生表情强度范围,从高维特征点轨迹中抽取一维的表情强度。最后,使用SVM建立表情模型和强度模型,进行表情的分类,并对高兴表情进行强度等级的归类。实验证明了该表情分析方法的有效性。
This paper does research on facial expression video sequences, recognizes facial expression and measures the intensity of expression based on a statistical learning method - support vector machine. First, an improved method of feature point tracking is used to extract the facial feature deformation. Then, a nonlinear dimensionality reduction method -ISOMAP is used to automatically generate the spectrum of expression intensity by extracting 1 - D manifold from high dimensional feature point trajectoties. At last, Support Vector Machine is used to build the expression model and intensity model to classify the facial expressions and classify the intensity grade of happy expression. Experiment has validated the effectiveness of this facial expression analysis method.
出处
《微处理机》
2007年第1期92-94,共3页
Microprocessors
基金
国家自然科学基金资助项目(60573079)
关键词
面部表情分析
特征点跟踪
支持向量机
表情强度
Facial expression analysis
Feature point tracking
Support Vector Machine
Expression intensity