Abstract: The paper regards flight maneuver evaluation as the problem of multivariate time series similarity measure，extracts the features of flight maneuver data based on SVD，and constructs the similarity matching model between two maneuvers.Using singular vector and eigenvector matrix as the features of flight maneuver not only has the results of dimension reduction and data denoising，but also unifies the flight maneuver data of different length to the same dimensionality.By the similarity measurement between under-evaluating flight maneuver and standard flight maneuver，it implements the quantitative evaluation of flight maneuvers.The experiments show good results on real data.

Key words: flight maneuver, flight data, Singular Value Decomposition（SVD）, feature extracting, similarity matching

摘要： 将飞行动作评价看成多元时间序列相似性匹配问题，提出了基于SVD的飞行动作数据特征提取方法，并构造了两个飞行动作的相似性匹配模型。将奇异值向量和特征矩阵作为飞行动作的特征，不仅具有数据降维和去噪的效果，并可将不同长度的飞行动作数据统一到同一个维度。通过对评估样本与标准样本之间的相似性度量，成功地实现了飞行动作的定量评价。真实数据上的实验表明该文的方法取得了较好的评价效果。

关键词: 飞行动作, 飞行数据, 奇异值分解, 特征提取, 相似性匹配