{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/shapedtw-shape-dynamic-time-warping","title":"shapeDTW: shape Dynamic Time Warping","arxiv_id":"1606.01601","date":"2016-06-06","proceeding":null,"authors":["Jiaping Zhao","Laurent Itti"],"abstract":"Dynamic Time Warping (DTW) is an algorithm to align temporal sequences with\npossible local non-linear distortions, and has been widely applied to audio,\nvideo and graphics data alignments. DTW is essentially a point-to-point\nmatching method under some boundary and temporal consistency constraints.\nAlthough DTW obtains a global optimal solution, it does not necessarily achieve\nlocally sensible matchings. Concretely, two temporal points with entirely\ndissimilar local structures may be matched by DTW. To address this problem, we\npropose an improved alignment algorithm, named shape Dynamic Time Warping\n(shapeDTW), which enhances DTW by taking point-wise local structural\ninformation into consideration. shapeDTW is inherently a DTW algorithm, but\nadditionally attempts to pair locally similar structures and to avoid matching\npoints with distinct neighborhood structures. We apply shapeDTW to align audio\nsignal pairs having ground-truth alignments, as well as artificially simulated\npairs of aligned sequences, and obtain quantitatively much lower alignment\nerrors than DTW and its two variants. When shapeDTW is used as a distance\nmeasure in a nearest neighbor classifier (NN-shapeDTW) to classify time series,\nit beats DTW on 64 out of 84 UCR time series datasets, with significantly\nimproved classification accuracies. By using a properly designed local\nstructure descriptor, shapeDTW improves accuracies by more than 10% on 18\ndatasets. To the best of our knowledge, shapeDTW is the first distance measure\nunder the nearest neighbor classifier scheme to significantly outperform DTW,\nwhich had been widely recognized as the best distance measure to date. Our code\nis publicly accessible at: https://github.com/jiapingz/shapeDTW.","url_abs":"http://arxiv.org/abs/1606.01601v1","url_pdf":"http://arxiv.org/pdf/1606.01601v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"shapedtw-shape-dynamic-time-warping","repo_url":"https://github.com/jiapingz/shapeDTW","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"none","reach":null},{"paper_slug":"shapedtw-shape-dynamic-time-warping","repo_url":"https://github.com/9552nZ/SmartSheetMusic","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok"}}],"tasks":[{"task_slug":"dynamic-time-warping","task_name":"Dynamic Time Warping"},{"task_slug":"temporal-sequences","task_name":"Temporal Sequences"},{"task_slug":"time-series-1","task_name":"Time Series"},{"task_slug":"time-series","task_name":"Time Series Analysis"}],"methods":[{"method_slug":"dtw","method_name":"DTW"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1606.01601","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}