Warping Algorithm Research Articles

Application of self-attention conditional deep convolutional generative adversarial networks in the fault diagnosis of planetary gearboxes

The Generative Adversarial Network (GAN) can generate samples similar to the original data to solve the problem of fault sample imbalance in planetary gearbox fault diagnosis. Most of models rely heavily on convolution to model the dependencies across feature vectors of vibration signals. However, the characterization ability of convolution operator is limited by the size of convolution kernel and it cannot capture the long-distance dependence in the original data. In this paper, self-attention is introduced into Conditional Deep Convolutional Generative Adversarial Networks (C-DCGAN). In the model, vibration features are dynamically weighted and merged, so that it can adaptively focus “attention” on different times to solve the problem of sample differences caused by time-varying vibration signals. Finally, the proposed method is verified on the planetary gearbox experiment and the quality of the generated signal samples is evaluated with Dynamic Time Warping (DTW) algorithm. The visual experimental results indicated that the proposed model performed better than conditional deep convolutional generative adversarial networks (C-DCGAN) and could accurately diagnose various working states of planetary gearboxes.

Research on Correlation Analysis Method of Time Series Features Based on Dynamic Time Warping Algorithm

Rich datasets related to the earth have been obtained because of the rapid development of earth observation technologies. A broad range of prior research has investigated how to obtain the correlation relationships of relevant features from a large number of data containing spatio-temporal information, which is also the technical basis for big data analysis. Based on the dynamic time warping (DTW) algorithm, this research proposes a correlation analysis method of time series data, and applies it to the correlation analysis between the time series features of the surface temperature and melting area of the Antarctic ice sheet. The results show that our method based on the DTW algorithm can effectively distinguish the change details of the non-linear time series. The correlation coefficients between these two highly-correlated factors computed by our method are higher than Pearson’s correlation coefficients by more than 0.2 almost in all study areas where data are available. In summary, the method proposed in this study provides a new feasible way for the correlation study of time-series data. It outperforms than traditional correlation coefficients such as Pearson’s correlation coefficient in some fields, specifically when complex nonlinear time series data with certain periodicity are used.

Learning an autoencoder to compress EEG signals via a neural network based approximation of DTW

In this paper, we propose a method to compress electroencephalography (EEG) data using a differentiable neural network trained with the dynamic time warping (DTW) algorithm. DTW is a popular method for aligning time series data, but it is computationally expensive and not differentiable, which makes it difficult to optimize using gradient-based methods. To circumvent this difficulty, we use a neural network based approximation of DTW as the reconstruction loss to learn a convolutional auto-encoder to compress EEG signals. We show that using a custom DTW as reconstruction loss for the autoencoder improves the sleep staging score of the reconstructed data, suggesting that it allows to better preserve important features of EEG data. We also show that these findings can be extended to multiple datasets and compression ratios.

Trajectory-Robust RFID Relative Localization Based on Phase Profile Correlation

Indoor localization systems based on radio frequency identification (RFID) technologies are widely adopted in the Internet of Things (IoT). Many of these applications prefer to obtain the relative position of the target object rather than the absolute position. Existing relative positioning methods are subject to strict trajectory constraints and cannot work under nonstraight trajectories. RFID mobile localization which has high accuracy can also be used for relative positioning, but they require the acquisition of precise moving trajectory coordinates. To overcome these challenges, we propose a trajectory-robust relative localization algorithm (TRL), based on phase profiles’ correlation in this article. The intuition of TRL is that the phase of the backscattered signal has a linear mapping relationship with the distance between the reader and the antenna, so the closer the tag is to each other, the higher the correlation of their phase curves is. We use the dynamic time warping (DTW) algorithm to calculate the correlation between the phase profiles of different tags and achieve relative localization of all the tags by sequentially determining their position intervals in the tag sequence. Compared with the state-of-the-art solutions, TRL gives greater flexibility to antenna movement and the demand for trajectory coordinates, and is it fully robust to trajectory error. A series of experimental results show that TRL has significant advantages over the current state-of-the-art schemes in terms of both accuracy and computation time.

Efficient real-time selective genome sequencing on resource-constrained devices.

Third-generation nanopore sequencers offer selective sequencing or "Read Until" that allows genomic reads to be analyzed in real time and abandoned halfway if not belonging to a genomic region of "interest." This selective sequencing opens the door to important applications such as rapid and low-cost genetic tests. The latency in analyzing should be as low as possible for selective sequencing to be effective so that unnecessary reads can be rejected as early as possible. However, existing methods that employ a subsequence dynamic time warping (sDTW) algorithm for this problem are too computationally intensive that a massive workstation with dozens of CPU cores still struggles to keep up with the data rate of a mobile phone-sized MinION sequencer. In this article, we present Hardware Accelerated Read Until (HARU), a resource-efficient hardware-software codesign-based method that exploits a low-cost and portable heterogeneous multiprocessor system-on-chip platform with on-chip field-programmable gate arrays (FPGA) to accelerate the sDTW-based Read Until algorithm. Experimental results show that HARU on a Xilinx FPGA embedded with a 4-core ARM processor is around 2.5× faster than a highly optimized multithreaded software version (around 85× faster than the existing unoptimized multithreaded software) running on a sophisticated server with a 36-core Intel Xeon processor for a SARS-CoV-2 dataset. The energy consumption of HARU is 2 orders of magnitudes lower than the same application executing on the 36-core server. HARU demonstrates that nanopore selective sequencing is possible on resource-constrained devices through rigorous hardware-software optimizations. The source code for the HARU sDTW module is available as open source at https://github.com/beebdev/HARU, and an example application that uses HARU is at https://github.com/beebdev/sigfish-haru.

Optimization of Characteristic Phenological Periods for Winter Wheat Extraction Using Remote Sensing in Plateau Valley Agricultural Areas in Hualong, China

It is important to develop or validate remote sensing methods to explore agricultural management and food self-sufficiency in the agricultural areas of the Qinghai–Tibet Plateau under the influence of global change, ecological protection, and socio-economic development. Studies on the use of remote sensing to monitor crop planting on the Qinghai-Tibetan Plateau are limited, with inconclusive results. Therefore, in this study, we analyzed Sentinel-2A/B images and field survey data in Hualong, China (located in Hehuang Valley, Qinghai-Tibetan Plateau) for winter wheat identification and verification at different spatial scales based on the time series of the normalized difference phenology index (NDPI) and dynamic time warping (DTW) algorithm. The characteristic phenological period and the corresponding DTW threshold were optimized using remote sensing data extracted for winter wheat. The results showed that NDPI corresponding to the jointing-heading stage, grouting-harvesting stage, and jointing-harvesting stage with DTW could identify winter wheat regardless of whether the spatial scale was a single quadrat, a combination of two quadrats, or the entire study area. The NDPI corresponding to the jointing-heading stage (corresponding DTW threshold T = 0.158) could generate a relatively rational winter wheat map; the NDPI corresponding to the time series of the grouting-harvesting stage (combined with DTW threshold T = 0.195) could detect a planting area with relatively high accuracy when supported by cultivated land, which matches the statistical reporting of the winter wheat area data. Similarly, with the support of cultivated land data, the planted area could be identified early based on the phenological characteristics of winter wheat before overwintering; however, the extraction scheme needs to be optimized further.

Fast economic dispatch with false data injection attack in electricity-gas cyber–physical system: A data-driven approach

For data-driven fast economic dispatch in the electricity-gas cyber–physical system (EGCPS), the strong reliance on communication networks makes it vulnerable to intentional cyberattacks. False data injection attack (FDIA) is emerging as a severe threat to secure system operation. To enhance the cybersecurity and computational efficiency in data-driven fast economic dispatch, a data-driven approach is proposed to serve both exact locational detection of FDIA and fast economic dispatch in real time. The data-driven approach concatenates a residual network (ResNet) and attention long short term memory model (ALSTM) based on LSTM and attention mechanism (called “ResNet-ALSTM”), which can achieve temporal correlations and feature extraction of large amounts of measurements. The proposed ResNet-ALSTM can serve as a multi-label classifier to detect the outlier locations of tampered measurements in the power system of the EGCPS. Further, the Fast Dynamic Time Warping algorithm can timely recover the tampered measurements of power system after FDIA. Finally, the proposed ResNet-ALSTM can achieve fast economic dispatch in the EGCPS with the recovered measurements of power system. The performances of the proposed methods are experimentally evaluated on IEEE 24-bus power system and Belgian 20-node gas system. The results demonstrate that the proposed methods can achieve the locational detection of FDIA, tampered data recovery and data-driven fast economic dispatch with high accuracy and computational efficiency. This research is a promising solution for safeguarding secure system operation of EGCPS against FDIA.

USING RVI AND NDVI TIME SERIES FOR CROPLAND MAPPING WITH TIME-WEIGHTED DYNAMIC TIME WARPING

Abstract. Monitoring and management of agricultural lands are essential due to reasons affecting agriculture, such as increasing population and global climate. With the increase in the temporal resolution of satellite systems, time-series classifications have become popular in cropland mapping. Because annual plants can give similar spectral reflectance values on the same date. In this context, agricultural land (∼500 km2) was selected in the south of South Dakota in the United States. The area includes alfalfa, corn, soybeans, winter wheat plants, developed, grassland/pasture, herbaceous wetlands, and open water areas. The study aims to map croplands with vegetation indices produced by annual Sentinel-1 and Sentinel-2 satellites. In this context, Radar Vegetation Index (RVI) produced from 25 Sentinel-1, and the Normalized Difference Vegetation Index (NDVI) produced from 26 Sentinel-2 satellites were used for 2020. We used the Time-Weighted Dynamic Time Warping (TWDTW) algorithm, which separates and classifies the similarities between two time series with variable speeds with time constraints. For mapping, the indices were classified both individually and combined. The highest overall accuracy (77.2%) was obtained with the combined use of NDVI and RVI. Among the plant classes, the lowest accuracy (83.71%) was found, and it was determined that the plant classes did not mix much. Sentinel-2 satellite is not available before April due to weather conditions in the region. For this reason, since the Sentinel-1 satellite is not affected by weather conditions, it is thought that the use of two satellites together will be beneficial in time series analysis.

A Sensor Fusion-Based GNSS Spoofing Attack Detection Framework for Autonomous Vehicles

This paper presents a sensor fusion-based Global Navigation Satellite System (GNSS) spoofing attack detection framework for autonomous vehicles (AVs) that consists of two strategies: (i) comparison between predicted location shift—i.e., distance traveled between two consecutive timestamps—and inertial sensor based location shift in addition to monitoring of vehicle motion states—i.e., standstill/ in motion; and (ii) detection and classification of turns (left or right) along with detection of vehicle motion states. In the first strategy, data from low-cost in-vehicle inertial sensors—i.e., speedometer, accelerometer, and steering angle sensor—are fused and fed to a long short-term memory (LSTM) neural network to predict the distance an AV will travel between two consecutive timestamps. The second strategy combines k-Nearest Neighbors (k-NN) and Dynamic Time Warping (DTW) algorithms to detect a turn and then classify left and right turns using steering angle sensor output. In both strategies, the GNSS-derived speed is compared with speedometer output to improve the effectiveness of the framework presented in this paper. To prove the efficacy of the sensor fusion-based attack detection framework, attack datasets are created for four unique spoofing attack scenarios—turn-by-turn, overshoot, wrong turn, and stop, using the publicly available real-world Honda Research Institute Driving Dataset (HDD). Analyses conducted in this study reveal that the sensor fusion-based detection framework successfully detects all four types of spoofing attacks within the required computational latency threshold.

Semi-supervised learning approach for localization and pose estimation of texture-less objects in cluttered scenes

3D object recognition and 6D pose estimation are crucial and fundamental endeavours for industrial assembly line automation such as robotic controlled pick-and-place. While the problem on textured objects is extensively studied, it is still an open research topic for texture-less industrial parts, e.g, solid cylinder and hollow tube, which are symmetric and appear similar in shapes from many viewing perspectives, causing pose ambiguity. Also, the industrial assembly line environment is usually cluttered and the captured data is noisy, which makes this task even more challenging. In this paper, we propose a novel object localization and pose estimation technique using RGB images and depth maps of industrial assembly parts. Our segmentation model is fully morphological and unsupervised for localizing the region of interest containing the target object extracted from the depth map. Our segmentation technique is effective in the presence of partial occlusion, multiple objects, and cluttered scenes. We use a model based approach for object recognition based on Stochastic Gradient Descent trained on features of Histogram of Oriented Gradients (HOG) and invariant moments of the region of interest containing the target object. We generate synthetic training images automatically from the CAD models of the industrial parts. We use a contour matching strategy based on Dynamic Time Warping (DTW) algorithm to estimate the optimal 6D pose of the object from a set of candidates. Experimental results show that our proposed approach competes and demonstrates advantages on the challenging T-LESS dataset compared to the state-of-the-art methods. • We introduce a strategy for generating multiple viewpoint synthetic images from a CAD model. • We introduce a non-supervised clustering process for object segmentation based on a depth map that does not require a training stage. • We demonstrate how different HOG feature parameters can be tuned to achieve better prediction and performance. • We present a new real-time strategy for object pose estimation using Dynamic Time Warping (DTW) between the target object contour and best prior candidates. • We show that our method achieves better time performance than other works for pose estimation.

Observation of Human Entrainment with Robot Assistance Towards Education

At present, the world faces the challenge of making education available to all the people of the world, especially in under-developed areas, where there is a lack of human resources. To improve educational standards and address this lack of human resources, we believe that robots can be a good teaching aid, to assist both the teacher and the student. This study investigated the entrainment phenomenon between human interactions with a robot as a third-party influence to provide learning packages that will be suited for students and will reduce the strain on teachers. We used dynamic time warping (DTW) algorithm in our time series analysis to effectively calculate the similarity of the brainwaves of the subjects. Then, we defined the entrainment phenomenon between the subjects. To ensure the complexity of this experiment, crosswords were used for this task, which subjects were required to complete in cooperation with each another. At different points of time, the robot would use different expressions, actions, and words as reminders or hints, aiding communication and better cooperation between subjects. The timing of the robotic reminders, synchronization, conversations between the subjects in the crossword tasks, and the completion rate of the crossword puzzles were recorded. The average completion rate improved by 10% with the aid of the robot, as opposed to not using the robot. The results of this study prove that robotic participation in human interaction is beneficial, and that implementing robotic assistance will improve education.

Fault location method combining transient mutation and dynamic time warping algorithm for grounding system

In the urban distribution network, due to the wide application of underground cables, the traditional resonant grounding system has gradually been unable to match the increasing fault current amplitude, real-time compensation and harmonic component compensation requirements. With the continuous advancement of reliable power electronics technology, flexible grounding systems (FGS) based on controllable converters are receiving more attention. Whether it is a traditional grounding system or a flexible grounding system, when a single-phase grounding fault occurs in the power grid, in order to ensure the safety of the power grid, it is necessary to accurately locate the fault in the system. However, how to apply traditional location methods to flexible grounding systems and high-resistance grounding faults is a challenge. In this paper, a new fault location method combining transient mutation and dynamic time warping (DTW) algorithm is proposed. The transient mutation is detected and used for the primary selection of the fault interval firstly. And then, the dynamic time warping algorithm determines the difference between the intervals, and the specific fault location is finally calculated and identified. The verification results based on MATLAB/Simulink prove the rationality and effectiveness of the proposed method.

Single-phase grounding fault line selection method based on dynamic time warping algorithm

Aiming at the problems of the complex line selection process and slow line selection speed of the existing intelligent classification line selection methods, a single-phase grounding fault line selection method based on the dynamic time warping (DTW) algorithm is proposed. The wavelet modulus maximum truncates the original signal to extract the temporary part. Then the DTW distance matrix between feeders is figured by the DTW algorithm. The fault line selection is realized by comparing the difference between the maximum, sub maximum, and minimum values with the threshold value. The simulation shows that the proposed approach can recognize proper line selection under various fault states and partial loss of recorded data, has high accuracy and has better robustness than the Euclidean distance and person correlation coefficient methods.

A Framework for Inclusion of Unmodelled Contact Tasks Dynamics in Industrial Robotics.

This paper presents a method to include unmodeled dynamics of load or a robot’s end-effector into algorithms for collision detection or general understanding of a robot’s operation context. The approach relies on the application of a previously developed modification of the Dynamic Time Warping algorithm, as well as a universally applicable algorithm for identifying kinematic parameters. The entire process can be applied to arbitrary robot configuration, and it does not require identification of dynamic parameters. The paper addresses the two main categories of contact tasks with unmodelled dynamics, which are determined based on whether the external contact force has a consistent profile in the end effector or base coordinate. Conclusions for representative examples analysed in the paper are applicable to tasks such as load manipulation, press bending, and crimping for the first type of forces and applications such as drilling, screwdriving, snap-fit, bolting, and riveting assembly for the latter category. The results presented in the paper are based on realistic testing with measurements obtained from an industrial robot.

Analysis and evaluation of hemiplegic gait based on wearable sensor network

Hemiplegia is a common symptom of acute cerebrovascular disease, and most patients with hemiplegia have abnormal gaits. Descriptive evaluation methods are commonly used in clinical for gait analysis, and outcomes are overly reliant on observation by rehabilitation physicians. The quantitative analysis of hemiplegia gait is urgently required to guide patients' rehabilitation training. This paper presents a quantitative analysis and evaluation method of hemiplegic gait based on inertial measurement units (IMUs). The wearable nodes are worn on the subjects’ waist and lower limbs to record data when they walked in a straight line. After the recorded data has been processed, the gradient descent algorithm (GDA) is used for attitude calculation, and the walking process of hemiplegic patients is reconstructed for gait analysis. Combined with kinematic analysis, three types of joint angles during walking are calculated, i.e., hip angle, knee angle, and ankle angle, and their comparison with the joint angles of normal gait is conducted. In terms of the joint angles, the phase variation of hemiplegic gait is analyzed first, then the gait difference between hemiplegic and normal subjects is measured by using a weighted dynamic time warping (WDTW) algorithm, and finally the gait distortion is evaluated quantitatively based on the WDTW distance. Experimental results demonstrate that the GDA-based gait reconstruction method and the WDTW-based gait evaluation method presented in this paper can quantify the abnormality of hemiplegic gait, and thereby monitor the rehabilitation process of patients' walking ability.

Dynamic Gesture Recognition of A-Mode Ultrasonic Based on the DTW Algorithm

At present, the research on gesture recognition based on ultrasonic signals is mostly about static gestures. Compared with static gestures, dynamic gestures reflect human intent more intuitively. However, the speed difference of dynamic gestures will cause a mismatch in time series. To tackle this problem, we proposed a dynamic gesture recognition scheme based on the dynamic time warping (DTW) algorithm. First, the ultrasonic signals of the forearm muscles were acquired by four A-mode ultrasonic transducers. Second, the original signals underwent Gaussian filtering, feature extraction, normalization, and principal component analysis (PCA) dimensionality reduction to obtain the main feature information and convert it into time series. Finally, the time series of dynamic gestures were matched based on the DTW algorithm, and the minimum Euclidean distance between different time series was calculated for prediction. We tested five dynamic gestures on ten subjects. The experiments were conducted for a total of ten rounds, and only one round was used as the training set. Results reveal that using the max feature can get an accuracy of 88.0% ± 5.9%. In addition, the energy feature (87.5% ± 6.4%) and the mean feature (85.2% ± 6.9%) also performed well. This research realizes the dynamic gesture recognition of A-mode ultrasonic for the first time, proving the potential of ultrasonic signals, and providing a wider range of use scenarios.

Accelerometer-based algorithm for the segmentation and classification of repetitive human movements during workouts

Monitoring a person’s physical activity has a wide range of applications in both sports and medicine. With the advancement of technology for measuring human movement, it is possible to monitor the performed activity without a need for an expert to directly overlook the trainee. While the initial interest focused mainly on aerobic exercises, research has recently begun to focus on strength exercises. The goal is to achieve the highest possible accuracy in tracking movement while maintaining the low cost and energy autonomy of the monitoring device. In this paper, an algorithm for the segmentation and classification of repetitive movements during workouts based on 3-axis accelerometer data from a wearable device is presented. The accelerometer signals were recorded continuously during the workout session which consisted typically of 9 strength exercises, where 8 default movements were repeated in three sets. Segmentation of the acceleration signals recorded during the workout was done using the frequency spectrum of the acceleration magnitude with an accuracy of 99.4%, while the classification of the segmented movements was done using the Dynamic Time Warping (DTW) algorithm with an accuracy of 85.7%.

Using Multi-Dimensional Dynamic Time Warping to Identify Time-Varying Lead-Lag Relationships.

This paper develops a multi-dimensional Dynamic Time Warping (DTW) algorithm to identify varying lead-lag relationships between two different time series. Specifically, this manuscript contributes to the literature by improving upon the use towards lead-lag estimation. Our two-step procedure computes the multi-dimensional DTW alignment with the aid of shapeDTW and then utilises the output to extract the estimated time-varying lead-lag relationship between the original time series. Next, our extensive simulation study analyses the performance of the algorithm compared to the state-of-the-art methods Thermal Optimal Path (TOP), Symmetric Thermal Optimal Path (TOPS), Rolling Cross-Correlation (RCC), Dynamic Time Warping (DTW), and Derivative Dynamic Time Warping (DDTW). We observe a strong outperformance of the algorithm regarding efficiency, robustness, and feasibility.

Matching method based on similarity of working trajectories

To identify whether the actual work trajectory of workers in the factory meets the predetermined work trajectory requirements, we proposed an efficient and accurate work trajectory similarity matching method. We comprehensively considered the similarity between the actual work track and the predetermined track from the two characteristics of track angle and track distance. Among them, the similarity of track rotation angle is calculated using the improved longest common subsequence algorithm, and the similarity of track distance is calculated using the improved dynamic time warping (DTW) algorithm. Then the results of the similarity calculation of these two features are weighted. Finally, the weighted results are used to evaluate the similarity between the actual work track and the predetermined track, so as to judge whether the actual work track meets the requirements of the predetermined track. Experimental data show that the trajectory similarity matching algorithm in this paper has higher accuracy and efficiency than traditional DTW and other algorithms, and has higher ability to resist the interference of trajectory point evacuation than traditional DTW and other algorithms.

Food price dynamics and regional clusters: machine learning analysis of egg prices in China

PurposeThe study uses machine learning techniques to cluster regional retail egg prices after 2000 in China. Furthermore, it combines machine learning results with econometric models to study determinants of cluster affiliation. Eggs are an inexpensiv, nutritious and sustainable animal food. Contextually, China is the largest country in the world in terms of both egg production and consumption. Regional clustering can help governments to imporve the precision of price policies and help producers make better investment decisions. The results are purely driven by data.Design/methodology/approachThe study introduces dynamic time warping (DTW) algorithm which takes into account time series properties to analyze provincial egg prices in China. The results are compared with several other algorithms, such as TADPole. DTW is superior, though it is computationally expensive. After the clustering, a multinomial logit model is run to study the determinants of cluster affiliation.FindingsThe study identified three clusters. The first cluster including 12 provinces and the second cluster including 2 provinces are the main egg production provinces and their neighboring provinces in China. The third cluster is mainly egg importing regions. Clusters 1 and 2 have higher price volatility. The authors confirm that due to transaction costs, the importing areas may have less price volatility.Practical implicationsThe machine learning techniques could help governments make more precise policies and help producers make better investment decisions.Originality/valueThis is the first paper to use machine learning techniques to cluster food prices. It also combines machine learning and econometric models to better study price dynamics.