Research of dynamic time warping approach based on path constraint

To solve computing complexity problems in some fields, dynamic time warping (DTW) based on path constraint is proposed. Piecewise-linear models are firstly built for the original patterns, and then a preliminary warping path is generated for the models. Based on the preliminary path, a global constraint is imposed on the warping path of the original patterns. Experimental results show that this algorithm improves computing efficiency remarkably with high accuracy in pattern classification. It can also produce warping path and traces for synchronizing the original patterns. The study in this paper is valuable to solve the problems on fault diagnosis of complex systems and processes.;