基于极大间隔最近邻学习的运动捕获数据检索

摘要

With the rapid development of human motion capture technology , large amounts of captured data has been gradually accumulated , the human motion retrieval technology becomes an essential key link in the process of motion data management and reuse . Logically similar motions may be numerically dissimilar , so it is difficult to get feasible results if the logical similarity between two movements is measured with Euclidean distance .Therefore, in this paper we present a novel approach for estimating the logical similarity between two motions with Mahalanobis distance metric between the motions , which is learned from large margin nearest neighbour metric learning algorithm, and then retrieve the motion .Experimental results show that method in this paper gets higher retrieval accuracy than the Euclidean distance and Linear Regression methods .%随着人体运动捕获技术的迅猛发展，逐渐积累了大量的捕获数据，人体运动检索技术成为运动数据管理和重用过程中必不可少的关键环节。由于逻辑相似的运动在数值上并不一定相似，使用欧式距离度量两个运动间的逻辑相似性难以取得理想的结果。为此，利用极大间隔最近邻度量学习算法，学习得到运动间的马氏距离度量，用以判断两个运动的逻辑相似性，进而进行运动的检索。实验结果表明，与欧式距离和线性回归相比，该方法能够获得更高的检索精度。