The well-known dynamic programming time warping algorithm (DTW) provides an optimal matching between 1-dimensional sequences in polynomial time. Finding an optimal 2-dimensional warping is an NP-complete problem. Hence, only approximate non-exponential time 2-dimensional warping algorithms currently exist. A polynomial time 2-dimensional approximation algorithm was proposed recently. This project provides a thorough analytical and experimental study of this algorithm. Its time complexity is improved from O(N^6) to O(N^4). An extension of the algorithm to 3D and potential higher-dimensional applications are described.
Worcester Polytechnic Institute
Major Qualifying Project
All authors have granted to WPI a nonexclusive royalty-free license to distribute copies of the work, subject to other agreements. Copyright is held by the author or authors, with all rights reserved, unless otherwise noted.