【摘要】 利用全局特征提取与局部特征提取相融合的方法,将傅里叶变换特征与方向梯度直方图特征相融合的算法结合支持向量机(SVM)来识别人体行为。首先将KTH和Weizmann数据集中提取出来的图像做预处理,分别提取出每张图像的傅里叶描述子和HOG特征,然后利用主成分分析法对HOG提取特征降维,最后用FD+HOG的融合特征放入SVM分类器中分类识别。实验结果表明,该算法识别率可达86%以上。更多还原

【Abstract】 Combing global with local feature extraction,we apply both Fourier transform features and direction gradient histogram into Support Vector Machine(SVM)to recognize human behavior.Images from the KTH and Weizmann data sets are processed to extract Fourier Descriptor(FD)and HOG feature of each image,and then principal component analysis is used to extract feature dimensionality from HOG.Finally,fusion features of FD+ HOG are put into the SVM classifier.Experimental results indicate that the recognition rate of the algorithm is greater than 86%.更多还原