首页>
外国专利>
Human Action Recognition Using Rreproducing Kernel Hilbert Space for Product manifold of Symmetric Positive definite Matrices
Human Action Recognition Using Rreproducing Kernel Hilbert Space for Product manifold of Symmetric Positive definite Matrices
展开▼
机译：对称正定矩阵乘积流的再现核希尔伯特空间的人类动作识别
展开▼
页面导航
摘要
著录项
相似文献
摘要
The present invention relates to a human behavior recognition method. More specifically, the present invention relates to a human behavior recognition method capable of recognizing human behavior by using a feature and a K-nearest neighbor algorithm in a reproducing kernel Hilbert space for a product manifold of a symmetric positive definite matrix. The method comprises the following steps of: receiving a test image and extracting a slope direction histogram feature vector of the test image (hereinafter, a first HOG feature vector) and an optical flow histogram feature vector (hereinafter, a first HOF feature vector); calculating a covariance descriptor matrix of the first HOG feature vector (hereinafter, a first test sample) and a covariance descriptor matrix of the first HOF feature vector (hereinafter, a second test sample); dimension-transforming the first test sample, the second test sample, a first training sample, and a second training sample into the reproducing kernel Hilbert space (RKHS), and calculating a distance between the first test sample and the second test sample and between the first training sample and the second training sample; and recognizing the human behavior of a training image in which the distance calculated by using the K-nearest neighbor algorithm is the closest as the human behavior of the test image.
展开▼
