关键词: 动作识别, 时间序列, 多元动态时间规整算法（MDTW）, 多维融合

Abstract: Scientific assessment of the quality of flight training is important to improve the driving skills of the pilots and flighting safety. Action recognition is an essential element of the flight assessment. Many assessments are based on the sequence of actions. This paper presents an MDTW（Multivariant Dynamic Time Warping） algorithm based on Dynamic Time Warping（DTW）. It conducts multivariate time series similarity computing by multidimensional integration. The results of experiments show the MDTW algorithm has improved efficiency and accuracy of conventional DTW algorithm.

Key words: action recognition, time series, Multivariant Dynamic Time Warping（MDTW）, multidimensional integration