Using Dynamic Time Warping to Find Patterns in Time Series

By D. Berndt, J. Clifford. 1994

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

A Comprehensive Survey on Transfer Learning (2019, methodology): “Specifically, the dynamic time warping algorithm [183] is applied to measuring similarity and finding the corresponding source-domain data point of each target-domain data point.”
Characteristic-Based Clustering for Time Series Data (2006, background): “Dynamic Time Warping (DTW) has been applied in time series mining to resolve the difficulty in clustering time series of variable lengths in Euclidean space or containing possible out-of-phase similarities (Berndt and Clifford, 1994; Ratanamahatana and Keogh, 2005).”
Time series shapelets: a novel technique that allows accurate, interpretable and fast classification (2010, background): “…(SAX) (Lin et al. 2007), and different similarity measures, such as Euclidean Distance (Faloutsos et al. 1994) and Dynamic Time Warping (DTW) (Berndt and Clifford 1994; Keogh and Ratanamahatana 2005), a recent large scale set of experiments indicates that the difference between different…”
Making Time-Series Classification More Accurate Using Learned Constraints (2004, methodology): “However, it has been empirically confirmed in many other settings, including music [20][47], finance [4], medicine [19], biometrics [18][34], chemistry [19], astronomy, robotics [42], and industry.”
Elastic functional coding of human actions: From vector-fields to latent variables (2015, methodology): “The most common way to solve for the mis-alignment problem is to use dynamic time warping (DTW) which originally found its use in speech processing [2].”
Experimental comparison of representation methods and distance measures for time series data (2010, methodology): “, Euclidean distance (ED) [18], Dynamic Time Warping (DTW) [10,35], distance based on Longest Common Subsequence (LCSS) [61], Edit Distance with Real Penalty (ERP) [13], Edit Distance on Real sequence (EDR) [14], DISSIM [20], Sequence Weighted Alignment model (Swale) [45], Spatial Assembling Distanc”
PDFormer: Propagation Delay-aware Dynamic Long-range Transformer for Traffic Flow Prediction (2023, methodology): “From the long-range view, we compute the similarity of the historical traffic flow between nodes using the Dynamic Time Warping (DTW) (Berndt and Clifford 1994) algorithm.”
Discovering Motor Programs by Recomposing Demonstrations (2020, methodology): “…our loss for a given demonstration τ obs is therefore characterized as: (2) As the trajectories τ obs , τ rec are possibly of different lengths, we use a pairwise matching cost between the two trajectories, where the optimal alignment is computed via dynamic time warping (Berndt & Clifford, 1994).”
Indexing multi-dimensional time-series with support for multiple distance measures (2003, methodology): “We describe an extension in 2D of the original DTW function as described by Berndt and Clifford [4].”
Discovering similar multidimensional trajectories (2002, methodology): “Berndt and Clifford [5] proposed to use this technique to measure the similarity of time-series data in data mining.”
Deep Learning for Intelligent Transportation Systems: A Survey of Emerging Trends (2020, background): “Dynamic Time Warping (DTW) [15] or the Longest Common Subsequence (LCSS) [96]), most maintain point-wise comparisons over the raw input space.”
A review on time series data mining (2011, methodology): “Based on the dynamic time warping (DTW) technique, the proposed method in (Berndt and Clifford, 1994) predefines some patterns to serve as templates for the purpose of pattern detection.”