최소 Frame차이를 이용한 Dynamic Time Warping에 관한 연구

홍광석 서울保健大學 1992 論文集 Vol.12 No.1

This paper reports on an optimum dynamic programming(DP) based time-normalization algorithm for spoken word recognition: The technique of dynamic programming for the time registration of a reference and a test pattern has found widespread use in the area of isolated word recognition. However, dynamic programming using spectrum distance measure has been exhausted processing time. In this paper, therefore, new warping constraint for decreasing processing time is proposed. The method is a linear path which are found method for beginning and ending point of reference and test patterns. The purpose of this investigation is to study the effects of such time saving on the performance of different dynamic time warping algorithm for a realistic speech database. The experiment shows that the present algorithm show performance comparable to or better than that of other dynamic time warping algorithm that were studied.

Dynamic Type-2 Fuzzy Time Warping (DT2FTW): A Hybrid Model for Uncertain Time-Series Prediction

Aref Safari,Rahil Hosseini,Mahdi Mazinani 한국지능시스템학회 2021 INTERNATIONAL JOURNAL of FUZZY LOGIC and INTELLIGE Vol.21 No.4

Prediction of time series is associated with nondeterministic pattern analysis for uncertain conditions. Therefore, it is necessary to develop high-quality prediction methods for real-world applications. Type-2 fuzzy systems can handle high-order uncertainties, such as sequential dependencies associated with time series. Precise and reliable prediction can help to develop reasonable strategies and assist specialists in planning the best policies for modeling events in uncertain time series. In this study, a hybrid model (dynamic type-2 fuzzy time warping [DT2FTW]) was proposed for handling high-order uncertainties in time-series prediction. A type-2 fuzzy intelligent system was developed alongside a dynamic time warping algorithm for predicting the patterns’ similarity in long-time series for time-series prediction. The results demonstrate that the proposed DT2FTW model yields more reliable predictions on global standard benchmarks such as the Mackey-Glass, Dow Jones, and NASDAQ time-series. The results also confirm that the proposed DT2FTW model has lower error rates than its counterpart algorithms in terms of the root mean square error (RMSE), mean absolute error (MAE), and mean percentage error (MPE). In addition, the results confirm the superiority of the proposed model with an average area under the ROC curve (AUC) of 94%, with the 95% confidence interval (92%-95%).

동적 시간워핑을 활용한 시계열자료의 군집분석

김성태,박만식 한국자료분석학회 2018 Journal of the Korean Data Analysis Society Vol.20 No.5

Two different approaches are considered for the clustering analysis of time-series data: time-domain approach and frequency-domain one. In the time domain, distance metrics measuring similarities among the time-series data take the estimation results under certain parametric models or autocorrelation structures inherent in each of the processes into account. The frequency-domain approach also plays an important role in time-series clustering analysis by transforming auto-covariance function into spectrum prior to measuring similarities among the processes. However, the previous time-series clustering approaches depend on assumptions of distribution or models. In this study, we apply the dynamic time warping (DTW) algorithm in which no assumptions are needed. This algorithm enables us to compare two time-series processes in order to measure similarities even when one process is temporally shifted from the other one. We evaluate the performance of DTW and compare with the metrics forementioned via the simulation study. For the real application, we considered the U.S. state-level seasonally adjusted monthly unemployment rate data. 시계열자료의 군집분석은 시간영역(time domain) 혹은 주파수영역(frequency domain)에서의 거리개념을 통해 이루어졌다. 시간 영역에서는 특정한 모수적(parametric) 모형을 적합한 후 모수 추정결과의 유사성을 고려하거나 자기상관구조(auto-correlation structure)의 유사성을 고려하여 거리개념을 도입하였다. 주파수 영역에서는 변동주기에 따른 자료의 순환구조를 의미하는 스펙트럼(spectrum)을 구한 후 적절한 변환을 통한 거리개념을 도입하였다. 본 논문에서는 주어진 원 시계열자료에 거리개념을 도입하되 동일한 시점 간의 거리 뿐 아니라 상이한 시점 간의 거리 또한 고려하는 동적 시간워핑(dynamic time warping; DTW)을 적용하고자 한다. 문자인식 및 행동인식 등의 여러 분야에서 활용되는 이 알고리즘은 시점에 국한하지 않은 측정값들 간의 비교를 가능케 한다. 모의실험을 통해 정상성 및 비정상성 하에서의 여러 시나리오 하에서, 시간영역과 주파수영역에서 널리 활용되는 다양한 거리들과 동적 시간워핑의 성능을 비교, 평가하였고 그 특성을 파악하였다. 또한 실증자료분석을 통해 미국 50개 주의 실업률 자료를 군집화하였고 동적 시간워핑방법을 이용하여 그 특성을 비교, 분석하였다.

2차 미분 연산자를 이용한 효과적인 Dynamic Time Warping

김세훈(Se-Hoon Kim),최형일(Hyung-Il Choi),이양원(Yang-Won Rhee),장석우(Seok-Woo Jang) 한국컴퓨터정보학회 2011 韓國컴퓨터情報學會論文誌 Vol.16 No.2

동적계획법이 기반인 Dynamic Time Warping은 두 패턴의 유사도를 비교하기 위해 널리 사용되는 방법이다. DTW알고리즘에는 2가지 알려진 문제점이 있다. 첫 번째 문제는 DTW알고리즘은 2개의 패턴의 대응경로를 계산하면서 특이점이 발생하는 문제가 발생하게 된다. 두 번째 문제는 동적패턴의 대응경로가 올바른지 알 수 없다는 문제가 있다. 이에 본 논문에서는 DTW알고리즘의 문제에 대한 효과적인 해결을 위하여, 2차 미분 연산자를 적용한 DTW알고리즘을 제안 한다. 2차 미분 연산자의 하나인 "라플라시안오브가우시안" 연산자를 적용하여, 효과적으로 특이점에 대한 문제를 해결하고, 올바른 대응경로를 가질 수 있는 방법에 대하여 제안하고, 실험의 결과로 제안하는 알고리즘의 우수성을 증명한다. Dynamic Time Warping based on Dynamic Programming is the one of the most widely been used to compare the similarity of two patterns. DTW algorithm has two known problems. The one is singularities. And the another problem is the accuracy of the warping path with patterns. Therefore, this paper suggest the solution for DTW algorithm to use a 2nd derivative operator. Laplacian of Gaussian is a kind of a 2nd derivative operator. Consequently, our suggestion method to apply to this operator, more efficient to solve the singularities problems and to secure a accuracy of the warping path. And the result shows a superior ability of this suggested method.

시계열 데이터 분류를 위한 고속 DTW 가속기 설계 및 구현

최원영(Wonyoung Choi),조재찬(Jaechan Cho),정윤호(Yunho Jung) 대한전자공학회 2021 전자공학회논문지 Vol.58 No.3

A Technology Analysis Model using Dynamic Time Warping

JunHyeog, Choi,SungHae, Jun 한국컴퓨터정보학회 2015 韓國컴퓨터情報學會論文誌 Vol.20 No.2

Technology analysis is to analyze technological data such as patent and paper for a given technology field. From the results of technology analysis, we can get novel knowledge for R&D planing and management. For the technology analysis, we can use diverse methods of statistics. Time series analysis is one of efficient approaches for technology analysis, because most technologies have researched and developed depended on time. So many technological data are time series. Time series data are occurred through time. In this paper, we propose a methodology of technology forecasting using the dynamic time warping (DTW) of time series analysis. To illustrate how to apply our methodology to real problem, we perform a case study of patent documents in target technology field. This research will contribute to R&D planning and technology management.

Anomaly Detection in Predictive Maintenance using Dynamic Time Warping

김영자,최경현 사단법인 한국융합기술연구학회 2024 아시아태평양융합연구교류논문지 Vol.10 No.1

Manufacturing systems face the fundamental challenge of efficient operation by leveraging vast amounts of real-time data collected through technological advancements such as artificial intelligence and machine learning. Maintenance systems have evolved to predict and manage equipment failures in advance, with data-driven fault detection being a crucial technology. However, most related research has been limited to single equipment for specific processes, making the direct application in actual manufacturing settings that use various equipment models or types challenging. When using multifacility models, the most crucial aspect is the analysis of variations and errors in the data collected from each facility. To mitigate the risk associated with a sole vendor, different models of equipment is used strategically, even for the same functionality. Consequently, collecting temporally mismatched data is prevalent. The current methodology, which has been predominantly focused on a single-facility approach, faces limitations in its application when dealing with unstructured, unlabeled data, or temporally mismatched data obtained across multiple facilities. This study employed the dynamic time warping (DTW) method to analyze discrepancies in time-series data obtained from multiple equipment groups by leveraging similarity analysis of data peak matching for anomaly detection. Specifically, an approach called auto time windowing is adopted to extract signal periods based on the detailed signal analysis results of the process, enabling the application of DTW. The auto time windowing allows for the accurate automated analysis of signal period by overcoming the limitations of analysis errors caused by noise in the existing data using the threshold of the actual signal. This methodology is validated for two different equipment groups involved in a real-world production process, where parts are attached to products. The results of this study demonstrated an improvement over conventional time-series analysis methods such as the Euclidean method, addressing errors that may occur. This research enhances the analysis theory using DTW for the actual problem of data discrepancies among multiple equipment groups in the manufacturing field, which is not previously considered in existing predictive maintenance (PdM) theories. This validation through case studies effectively contributes to expanding the utilization of PdM.

Weighted Dynamic Time Warping을 이용한 비행기동 패턴인식 알고리즘

이경선,이찬석,박상선,김태욱,조인제 한국항공우주학회 2015 한국항공우주학회 학술발표회 논문집 Vol.2015 No.11

본 논문에서는 초음속 항공기의 비행시험 데이터를 이용해 비행기동 패턴을 인식하여 특정한 기동구간을 자동으로 검출하는 알고리즘에 대해 제안한다. 비행기동 패턴은 대표적으로 Pitch Doublet Sequence, Roll Doublet Sequence, Yaw Doublet Sequence, Wings Level Sideslip 등으로 정의하고, 각각의 비행기동 패턴을 인식하기 위해 템플릿매칭(Template Matching) 기반의 접근법인 Weighted Dynamic Time Warping(WDTW) 알고리즘을 사용한다. WDTW 알고리즘은 제스처, 음성인식 등 다양한 분야에 널리 사용되는 접근법으로 적은 수의 훈련데이터로 높은 인식률을 나타내어 비행기동 패턴인식을 위한 효과적인 알고리즘이다. 비행시험 데이터 특징 분석 및 비행기동 검출 실험결과를 통해 제안한 알고리즘의 효용성을 입증한다. In this paper, flight maneuver pattern recognition algorithm is proposed using Weighted Dynamic Time Warping (WDTW). In conventional flight analysis, flight maneuvers are manually marked point by point. Due to this fact, a lot of time and effort are required for finding the maneuvers. To mitigate this problem, an appropriate framework based on WDTW is developed. The objective of this paper, therefore, is to recognize pattern of the maneuvers automatically with an adaptive weight based distance measure. The maneuvers are defined as Pitch doublet sequence, Roll doublet sequence, Yaw doublet sequence, and Wings level sideslip from a supersonic jet plane. The method was evaluated on a set of flight data. As documented by the experimental results, the proposed method offers a promising result in terms of detection probability and false alarm rate.

고랭지배추 무게와 생육환경변수 간의 임의적인 시차 관계 및 왜곡에 관한 연구

이윤숙 한국농업경제학회 2022 農業經濟硏究 Vol.63 No.2

The study analyzed dynamic time relationships and time warping between weight of highland chinese cabbage and growth environment variables using an arbitrary-lag causality method. During an early stage of growth, the relation between cabbage and growth environment variables is formed in arbitrary times. The bigdata used in the study are collected from the open field samrtfarm system operated by Rural Development Administration. The study applied for a dynamic time warping algorithm to measure dynamic time distances between weights of highland chinese cabbage and growth and weather variables, respectively. After that, we compared arbitrary-lag causality and fixed-lag causality. We found the existence of arbitrary-time lags between variables. In addiction, we found that the arbitrary-lag causality analysis is performed better than fixed-lag causality. Finally, we found that growth and weather variables caused cabbage weight under the arbitray-time lags, but not vice versa.

HybridFTW를 사용한 효율적인 k-NN 검색

이민우(Minwoo Lee),문양세(Yang-Sae Moon),김진호(Jinho Kim) 한국정보과학회 2013 정보과학회논문지 : 데이타베이스 Vol.40 No.6

시계열 데이터를 대상으로 한 유사 검색에서 동적 타임 워핑(dynamic time warping: DTW) 거리를 효율적으로 계산하기 위해 많은 연구가 수행되었다. DTW 거리는 유사 검색에서 높은 정확도를 제공하지만, 계산이 복잡하여 대용량 데이터베이스에 적용하기 힘든 문제점이 있다. 이러한 DTW 거리의 효율적 계산 방법으로 FastDTW와 FTW가 제안되었고, 최근 이 두 방법의 장점을 취한 HybridFTW가 제안되었다. HybridFTW는 FastDTW의 허용 범위 제한을 통한 빠른 계산의 장점과 FTW의 미리 버림의 장점을 조합한 하이브리드 접근법이다. 본 논문에서는 우선 FTW와 FastDTW의 장점을 각각 분석하고, 이들의 장점을 취한 HybridFTW의 개념을 설명한다. 그런 다음, HybridFTW를 k-NN 검색에 사용하기 위한 주요 절차를 다섯 단계로 나누어 설명한다. 그리고, 이들 다섯 단계를 사용하여 HybridFTW 기반의 k-NN알고리즘을 새롭게 제안하고, 그 정확성을 정리로서 제시하고 증명한다. 마지막으로, 실제 실험을 통해 제안한 알고리즘이 기존의 FastDTW와 FTW를 기반한 알고리즘보다 최대 20배까지 성능을 향상시킴을 보인다. There have been many research efforts on computing the DTW(dynamic time warping) distance efficiently in similarity search on time-series databases. The DTW distance is known to offer the high accuracy in similarity search, but it has a critical problem in supporting the large database due to its high computation complexity. For the fast computation of the DTW distance, FastDTW and FTW have been proposed recently, and HybridFTW has also been proposed to adopt both of their advantages. HybridFTW is a hybrid approach that combines the advantage of FastDTW, which provides the fast computation through the limitation of allowable ranges, and the advantage of FTW, which exploits the early abandon effect. In this paper, we first analyze the computation procedure of FastDTW and FTW in detail and present the concept of HybridFTW by taking both of their advantages. After then, we propose a HybridFTW-based k-NN algorithm. For this, we first explain five major steps to implement the HybridFTW-based k-NN search in detail. We next propose a formal k-NN algorithm exploiting HybridFTW and prove its correctness through a formal theorem. Experimental results for real and synthetic data sets show that the proposed k-NN algorithm improves the search performance by up to 20 times over k-NN algorithms based on FastDTW and FTW.