基于最长公共子序列的人体运动序列检索

Retrieval of Human Motion Capture Data Based on Longest Common Subsequence Model

在计算机仿真领域,运用重新组合等相关技术,可以对已经捕捉的人体运动数据实现重用,以较低的成本产生新的运动。但是,由于运动库的数据庞大,对其实施有效检索成为一个重要问题。以倒排表数据结构为基础,设计出一套针对人体运动的检索算法。不同于传统的最长公共子序列(Longest Common Subsequence,LCSS)的度量算法,提出了一种限制最小匹配率ρmin的有限最长公共子序列(Limited-LCSS)算法,并在此基础上针对倒排表的数据结构特点对算法进行了优化,显著的提高了算法的效率。实验表明提出的检索方法具有较好的速度和准确性。

In the field of the Computer Emulation, the captured human sports data can be reused with the recombination, which can generate new sports with lower cost. However, it is important to research on efficiency of retrieval method due to huge data. A method of researching in the human sports data was advanced with the inverted index as data structure and Limited-LCSS as the measure, which improved the original LCSS algorithm by define Pmin as the minimal matching rate and using characteristic of inverted index model The experiment shows that the method is accurate and efficient.