Dynamic Time Warping Measure Research Articles

Multimodal Satellite Image Time Series Analysis Using GAN-Based Domain Translation and Matrix Profile

The technological development of the remote sensing domain led to the acquisition of satellite image time series (SITS) for Earth Observation (EO) by a variety of sensors. The variability in terms of the characteristics of the satellite sensors requires the existence of algorithms that allow the integration of multiple modalities and the identification of anomalous spatio-temporal evolutions caused by natural hazards. The unsupervised analysis of multimodal SITS proposed in this paper follows a two-step methodology: (i) inter-modality translation and (ii) the identification of anomalies in a change-detection framework. Inter-modality translation is achieved by means of a Generative Adversarial Network (GAN) architecture, whereas, for the identification of anomalies caused by natural hazards, we adapt the task to a similarity search in SITS. In this regard, we provide an extension of the matrix profile concept, which represents an answer to identifying differences and to discovering novelties in time series. Furthermore, the proposed inter-modality translation allows the usage of standard unsupervised clustering approaches (e.g., K-means using the Dynamic Time Warping measure) for mono-modal SITS analysis. The effectiveness of the proposed methodology is shown in two use-case scenarios, namely flooding and landslide events, for which a joint acquisition of Sentinel-1 and Sentinel-2 images is performed.

Fitting nonlinear models to continuous oxygen data with oscillatory signal variations via a loss based on Dynamic Time Warping*

High throughput experimental systems play an important role in bioprocess development, as they provide an efficient way of analysing different experimental conditions and perform strain discrimination in previous phases to the industrial scale production. In the millilitre scale, these systems are combinations of parallel mini-bioreactors, liquid handling robots and automated workflows for data handling and model based operation. For successfully monitoring cultivation conditions and improving the overall process quality by model-based approaches, a proper model identification is crucial. However, the quality and amount of measurements makes this task challenging considering the complexity of the bio-processes. The Dissolved Oxygen Tension is often the only measurement which is available online, and therefore, a good understanding of the errors in this signal is important for performing a robust estimation. Some of the expected errors will provoke uncertainties in the time-domain of the measurement, and in those cases, the common Weighted Least Squares estimation procedure can fail providing good results. Moreover, these errors will have even a larger effect in the fed-batch phase where bolus feeding is applied, as this generates fast dynamic responses in the signal. In the present work, an insilico study of the performance of Weighted Least Squares estimator is analysed when the expected time-uncertainties are present in the oxygen signal. As an alternative, a loss based on the Dynamic Time Warping measure is proposed. The results show how this latter procedure outperforms the former reconstructing the oxygen signal, and in addition, returns less biased parameter estimates.

Weighted dynamic time warping for traffic flow clustering

This paper presents a novel similarity measure to identify interesting traffic patterns on a large traffic flow time series data for the central suburbs of Melbourne city in Australia. This new measure is a weighted Dynamic Time Warping (DTW) method based on Gaussian probability function, named GWDTW, that reflects the relative importance of peak hours. We have shown its superior performance over two benchmark similarity measures, the Euclidean distance and conventional DTW measure, on the intersection clustering task using k-medoids clustering algorithm, with respect to both internal and external evaluation measures. With intensive evaluation, the results show that GWDTW is a very effective similarity measure for modelling traffic behaviours, which can provide policy makers with more valuable information for infrastructure design, and smart city development.

Recognition of Human Activities Using Depth Maps and the Viewpoint Feature Histogram Descriptor.

In this paper we propose a way of using depth maps transformed into 3D point clouds to classify human activities. The activities are described as time sequences of feature vectors based on the Viewpoint Feature Histogram descriptor (VFH) computed using the Point Cloud Library. Recognition is performed by two types of classifiers: (i) k-NN nearest neighbors’ classifier with Dynamic Time Warping measure, (ii) bidirectional long short-term memory (BiLSTM) deep learning networks. Reduction of classification time for the k-NN by introducing a two tier model and improvement of BiLSTM-based classification via transfer learning and combining multiple networks by fuzzy integral are discussed. Our classification results obtained on two representative datasets: University of Texas at Dallas Multimodal Human Action Dataset and Mining Software Repositories Action 3D Dataset are comparable or better than the current state of the art.

ASSESSING DYNAMIC-TIME-WARPING DISSIMILARITY MEASURES IN REGIONALIZATION OF RIVER DISCHARGES

Regionalization of river discharges is a process of transferring hydrological information to generalize hydrological information from one river to another. One approach to regionalize river discharge is to use a distance-based regional analysis by employing a Dynamic Time Warping (DTW) dissimilarity measure to cluster homogeneous river discharge patterns based on sequenced of time series discharge data. However, clustering homogeneous river discharge patterns can be sensitive to the choice of distance metric measures used due to out of phase behavior in the discharge time series. In this study, we assess three types of Dynamic Time Warping (DTW) measures specifically conventional DTW, a feature based DTW and a weighted based DTW on four annual discharge time series from six rivers in the state of Johor, Malaysia. A comparison of eight different clustering validation indices to determine the optimal number of rivers clusters with similar discharge patterns. These indices are used to measure the internal and external strength of the identified clusters. The results indicate that weighted based DTW outperform the conventional DTW and feature based DTW with 75% of the clustering indices agree that there are three optimal clusters of river discharge. By using weight as a function in DTW, it helps to cater the out of phase behavior in river discharge time series with the highest agreement of clustering indices compared to other types of DTW measures. We also found that three of the rivers (Sayong, Bekok, and Segamat) have similar river discharge patterns and could be used together in the generalization process. Meanwhile, the other rivers (Johor, Kahang, and Muar) varies in their time series patterns.

Dynamic Time Warping Dissimilarity Matrices

The dynamic time-warping measure (DTW), also known as the DTW distance, is a widely used method that quantifies how dissimilar two time series are from each other. That quantification of the degree of dissimilarity is often motivated by tasks such as data exploration/visualization, clustering, and classification.

Learning a Mahalanobis Distance-Based Dynamic Time Warping Measure for Multivariate Time Series Classification.

Multivariate time series (MTS) datasets broadly exist in numerous fields, including health care, multimedia, finance, and biometrics. How to classify MTS accurately has become a hot research topic since it is an important element in many computer vision and pattern recognition applications. In this paper, we propose a Mahalanobis distance-based dynamic time warping (DTW) measure for MTS classification. The Mahalanobis distance builds an accurate relationship between each variable and its corresponding category. It is utilized to calculate the local distance between vectors in MTS. Then we use DTW to align those MTS which are out of synchronization or with different lengths. After that, how to learn an accurate Mahalanobis distance function becomes another key problem. This paper establishes a LogDet divergence-based metric learning with triplet constraint model which can learn Mahalanobis matrix with high precision and robustness. Furthermore, the proposed method is applied on nine MTS datasets selected from the University of California, Irvine machine learning repository and Robert T. Olszewski's homepage, and the results demonstrate the improved performance of the proposed approach.

Early recognition of upper limb motor tasks through accelerometers: real-time implementation of a DTW-based algorithm

A new real-time implementation of a Dynamic Time Warping (DTW)-based classification scheme is presented here, and its performance evaluated on experimental data. Nine young adults were requested to perform instances of eight different purposeful movements described in the Wolf Motor Function Test, while wearing a three-axis accelerometer sensor placed on the inner forearm. Results include the correct recognition percentage, as compared to a classification scheme based on the traditional DTW measure, and the recognition percentage as a function of the time elapsed from the beginning of the performed movements. The Real-Time DTW basically performs with the same accuracy of the traditional DTW-based classification scheme (91.5% of correct recognition percentage), a figure that increases to 96.5% if the multidimensional scheme is adopted. Moreover, more than 60% of movements are correctly recognized before their end, thus setting the way for applications in rehabilitation and assistive technologies, where a real-time control scheme is able to interact with the user while the movement is being performed.

Curve Matching, Time Warping, and Light Fields: New Algorithms for Computing Similarity between Curves

The problem of curve matching appears in many application domains, like time series analysis, shape matching, speech recognition, and signature verification, among others. Curve matching has been studied extensively by computational geometers, and many measures of similarity have been examined, among them being the Frechet distance (sometimes referred in folklore as the "dog-man" distance). A measure that is very closely related to the Frechet distance but has never been studied in a geometric context is the Dynamic Time Warping measure (DTW), first used in the context of speech recognition. This measure is ubiquitous across different domains, a surprising fact because notions of similarity usually vary significantly depending on the application. However, this measure suffers from some drawbacks, most importantly the fact that it is defined between sequences of points rather than curves. Thus, the way in which a curve is sampled to yield such a sequence can dramatically affect the quality of the result. Some attempts have been made to generalize the DTW to continuous domains, but the resulting algorithms have exponential complexity. In this paper we propose similarity measures that attempt to capture the "spirit" of dynamic time warping while being defined over continuous domains, and present efficient algorithms for computing them. Our formulation leads to a very interesting connection with finding short paths in a combinatorial manifold defined on the input chains, and in a deeper sense relates to the way light travels in a medium of variable refractivity.