shapeDTW: Shape Dynamic Time Warping

作者：

Highlights：

• Developed an improved sequence alignment algorithm, named shapeDTW, which augments the traditional Dynamic Time Warping (DTW) by local temporal shape information.

• shapeDTW performs significantly better than DTW under the one nearest neighbor classifier for time series classification; to be concrete, it wins DTW on 64 (out of 84) UCR time series datasets.

• shapeDTW is essentially a DTW algorithm, therefore runs efficiently. Moreover, shapeDTW is insensitive to one design parameter.

摘要

•Developed an improved sequence alignment algorithm, named shapeDTW, which augments the traditional Dynamic Time Warping (DTW) by local temporal shape information.•shapeDTW performs significantly better than DTW under the one nearest neighbor classifier for time series classification; to be concrete, it wins DTW on 64 (out of 84) UCR time series datasets.•shapeDTW is essentially a DTW algorithm, therefore runs efficiently. Moreover, shapeDTW is insensitive to one design parameter.

论文关键词：Dynamic Time Warping,Sequence alignment,Time series classification

论文评审过程：Received 17 October 2016, Revised 5 August 2017, Accepted 12 September 2017, Available online 14 September 2017, Version of Record 22 September 2017.

论文官网地址：https://doi.org/10.1016/j.patcog.2017.09.020