Flight Motion Recognition Method Based on Multivariate Phase Space Reconstruction and Approximate Entropy

Abstract

In this paper, the Flight Motion Recognition Method (FMRM) based on Multivariate Phase Space Reconstruction (MPSR) and Approximate Entropy (ApEn) is proposed against the problem of complex rule representation and costly training in the current research of Flight Motion Recognition (FMR). Firstly, the cardinality of Phase Space Reconstruction (PSR) is unified, and the PSR sequence of height, roll angle and pitch angle is obtained respectively. Then, phase points at the same position of the three reconstructed sequences is merged based on Bayesian theory to generate the new fusion sequence where the maneuvering motion segment is extracted. Finally, various flight actions can be identified according to the numerical difference in sequence complexity of action segments which is qualitatively and quantitatively described by Recurrence Plot (RP) and ApEn. Verifying the effectiveness of FMRM through multiple real continuous flight data. The result shows that the FMRM can quickly and accurately identify different maneuvers, which lays good foundation for the follow-up evaluation.