Flight Training Subject Identification Method Based on Multivariate Subsequence Search With Double Windows

Abstract

To solve the problem of low completeness in identifying flight training subjects in flight training data, a Multivariate Subsequence Search with Double Windows (MSDW) algorithm is proposed based on the Euclidean distance. First, a method for retrieving locally optimal subsequences is proposed based on constructing a distance matrix of candidate subsequences; then, the method is applied to the MSDW algorithm for retrieving all eligible locally optimal subsequences; finally, a double windows method is employed to improve the completeness of matched subsequences. Theoretical analysis and experimental results show that the MSDW algorithm is well designed and can correctly and effectively retrieve the local optimal subsequence, which can achieve recognition accuracy of 1 in the engineering application of single flight training subject recognition. It is experimentally verified that compared with the traditional SPRING algorithm, the completeness of the MSDW algorithm in recognizing individual flight training subjects is greatly improved, and the effectiveness of flight training subject recognition is effectively enhanced.

Keywords

• Dynamic time warping
• euclidean distance
• flight training subject identification
• multivariate time series
• subsequence similarity retrieval