Dynamic time warping

About: Dynamic time warping is a research topic. Over the lifetime, 6013 publications have been published within this topic receiving 133130 citations.

Using time-series similarity measures to compare animal movement trajectories in ecology

Abstract: Identifying and understanding patterns in movement data are amongst the principal aims of movement ecology. By quantifying the similarity of movement trajectories, inferences can be made about diverse processes, ranging from individual specialisation to the ontogeny of foraging strategies. Movement analysis is not unique to ecology however, and methods for estimating the similarity of movement trajectories have been developed in other fields but are currently under-utilised by ecologists. Here, we introduce five commonly used measures of trajectory similarity: dynamic time warping (DTW), longest common subsequence (LCSS), edit distance for real sequences (EDR), Frechet distance and nearest neighbour distance (NND), of which only NND is routinely used by ecologists. We investigate the performance of each of these measures by simulating movement trajectories using an Ornstein-Uhlenbeck (OU) model in which we varied the following parameters: (1) the point of attraction, (2) the strength of attraction to this point and (3) the noise or volatility added to the movement process in order to determine which measures were most responsive to such changes. In addition, we demonstrate how these measures can be applied using movement trajectories of breeding northern gannets (Morus bassanus) by performing trajectory clustering on a large ecological dataset. Simulations showed that DTW and Frechet distance were most responsive to changes in movement parameters and were able to distinguish between all the different parameter combinations we trialled. In contrast, NND was the least sensitive measure trialled. When applied to our gannet dataset, the five similarity measures were highly correlated despite differences in their underlying calculation. Clustering of trajectories within and across individuals allowed us to easily visualise and compare patterns of space use over time across a large dataset. Trajectory clusters reflected the bearing on which birds departed the colony and highlighted the use of well-known bathymetric features. As both the volume of movement data and the need to quantify similarity amongst animal trajectories grow, the measures described here and the bridge they provide to other fields of research will become increasingly useful in ecology.

Velocity and acceleration features in speaker recognition

Abstract: The performance of dynamic features in automatic speaker recognition is examined. Second- and third-order regression analysis examining the performance of the associated feature sets independently, in combination, and in the presence of noise is included. It is shown that each regression order has a clear optimum. These are independent of the analysis order of the static feature from which the dynamic features are derived, and insensitive to low-level noise added to the test speech. It is also demonstrated that while the static feature gives the best individual performance, multiple linear combinations of feature sets based on regression analysis can reduce error rates. >

Dynamic-Time-Warping-Based Measurement Data Alignment Model for Condition-Based Railroad Track Maintenance

Abstract: Condition-based maintenance is believed to be a cost-effective and safety-assured strategy for railroad track management. Implementation of the strategy strongly relies on reliable and complete track condition data, reliable track deterioration models, and efficient and solvable mathematical models for optimal track maintenance scheduling. In practice, reliability of track condition inspection data is often in question; therefore, collected inspection data need to be preprocessed before it is used to implement a condition-based maintenance strategy. Reliable track condition inspection data means accurate positioning data and noiseless condition parameter measurements. Based on dynamic time warping, which is a widely used technique in the area of speech signal processing and biomedical engineering, this paper presents a robust optimization model for correcting positional errors of inspection data from a track geometry car, which is a kind of specialized instrument that is extensively used to measure the condition of tracks under wheel loadings. An efficient solution algorithm for the model is proposed as well. Applications of the model to inspection data from the track geometry car show that positional errors are almost removed from the inspection data, regardless of noises in condition parameter measurements and track maintenance interventions, and the model takes 1.5004 s, on average, to complete the positional error correction for a 1-km-long track segment. The presented model is adjustable to alignment of data sequences in many other areas, e.g., railroad inspection by track geometry trolley, highway roughness inspection by Light Detection and Ranging (LiDAR) vehicles, and railroad catenary wire geometry inspection.

Dynamic Image-to-Class Warping for Occluded Face Recognition

Abstract: Face recognition (FR) systems in real-world applications need to deal with a wide range of interferences, such as occlusions and disguises in face images. Compared with other forms of interferences such as nonuniform illumination and pose changes, face with occlusions has not attracted enough attention yet. A novel approach, coined dynamic image-to-class warping (DICW), is proposed in this work to deal with this challenge in FR. The face consists of the forehead, eyes, nose, mouth, and chin in a natural order and this order does not change despite occlusions. Thus, a face image is partitioned into patches, which are then concatenated in the raster scan order to form an ordered sequence. Considering this order information, DICW computes the image-to-class distance between a query face and those of an enrolled subject by finding the optimal alignment between the query sequence and all sequences of that subject along both the time dimension and within-class dimension. Unlike most existing methods, our method is able to deal with occlusions which exist in both gallery and probe images. Extensive experiments on public face databases with various types of occlusions have confirmed the effectiveness of the proposed method.