• Journal of KIISE:Databases
• /
• v.30 no.4
• /
• pp.381-396
• /
• 2003

Subsequence matching is an operation that finds subsequences whose changing patterns are similar to a given query sequence from time-series databases. This paper points out the performance bottleneck in subsequence matching, and then proposes an effective method that improves the performance of entire subsequence matching significantly by resolving the performance bottleneck. First, we analyze the disk access and CPU processing times required during the index searching and post processing steps through preliminary experiments. Based on their results, we show that the post processing step is the main performance bottleneck in subsequence matching, and them claim that its optimization is a crucial issue overlooked in previous approaches. In order to resolve the performance bottleneck, we propose a simple but quite effective method that processes the post processing step in the optimal way. By rearranging the order of candidate subsequences to be compared with a query sequence, our method completely eliminates the redundancy of disk accesses and CPU processing occurred in the post processing step. We formally prove that our method is optimal and also does not incur any false dismissal. We show the effectiveness of our method by extensive experiments. The results show that our method achieves significant speed-up in the post processing step 3.91 to 9.42 times when using a data set of real-world stock sequences and 4.97 to 5.61 times when using data sets of a large volume of synthetic sequences. Also, the results show that our method reduces the weight of the post processing step in entire subsequence matching from about 90% to less than 70%. This implies that our method successfully resolves th performance bottleneck in subsequence matching. As a result, our method provides excellent performance in entire subsequence matching. The experimental results reveal that it is 3.05 to 5.60 times faster when using a data set of real-world stock sequences and 3.68 to 4.21 times faster when using data sets of a large volume of synthetic sequences compared with the previous one.

• Journal of KIISE:Databases
• /
• v.34 no.6
• /
• pp.503-517
• /
• 2007

In this paper we address the MBR-safe property of Piecewise Aggregation Approximation(PAA), and propose an of efficient subsequence matching method based on the MBR-safe PAA. A transformation is said to be MBR-safe if a low-dimensional MBR to which a high- dimensional MBR is transformed by the transformation contains every individual low-dimensional sequence to which a high-dimensional sequence is transformed. Using an MBR-safe transformation we can reduce the number of lower-dimensional transformations required in similar sequence matching, since it transforms a high-dimensional MBR itself to a low-dimensional MBR directly. Furthermore, PAA is known as an excellent lower-dimensional transformation single its computation is very simple, and its performance is superior to other transformations. Thus, to integrate these advantages of PAA and MBR-safeness, we first formally confirm the MBR-safe property of PAA, and then improve subsequence matching performance using the MBR-safe PAA. Contributions of the paper can be summarized as follows. First, we propose a PAA-based MBR-safe transformation, called mbrPAA, and formally prove the MBR-safeness of mbrPAA. Second, we propose an mbrPAA-based subsequence matching method, and formally prove its correctness of the proposed method. Third, we present the notion of entry reuse property, and by using the property, we propose an efficient method of constructing high-dimensional MBRs in subsequence matching. Fourth, we show the superiority of mbrPAA through extensive experiments. Experimental results show that, compared with the previous approach, our mbrPAA is 24.2 times faster in the low-dimensional MBR construction and improves subsequence matching performance by up to 65.9%.

• Korean Journal of Remote Sensing
• /
• v.38 no.5_1
• /
• pp.597-608
• /
• 2022

Remotely sensed data have been used in various fields, such as disasters, agriculture, urban planning, and the military. Recently, the demand for the multitemporal dataset with the high-spatial-resolution has increased. This manuscript proposed an automatic image matching algorithm using a deep learning technique to utilize a multitemporal remotely sensed dataset. The proposed deep learning model was based on High Resolution Net (HRNet), widely used in image segmentation. In this manuscript, denseblock was added to calculate the correlation map between images effectively and to increase learning efficiency. The training of the proposed model was performed using the multitemporal orthophotos of the National Geographic Information Institute (NGII). In order to evaluate the performance of image matching using a deep learning model, a comparative evaluation was performed. As a result of the experiment, the average horizontal error of the proposed algorithm based on 80% of the image matching rate was 3 pixels. At the same time, that of the Zero Normalized Cross-Correlation (ZNCC) was 25 pixels. In particular, it was confirmed that the proposed method is effective even in mountainous and farmland areas where the image changes according to vegetation growth. Therefore, it is expected that the proposed deep learning algorithm can perform relative image registration and image matching of a multitemporal remote sensed dataset.

• Proceedings of the Korean Information Science Society Conference
• /
• 2002.04b
• /
• pp.124-126
• /
• 2002

본 논문에서는 비디오 데이터가 지니는 이동 객체의 움직임 궤적(moving objects'trajectories)에 대해 유사 부분 움직임 궤적 검색을 효율적으로 지원하는 N-워핑(N-warping) 알고리즘을 제안한다. 제안하는 알고리즘은 기존의 시계열 데이터베이스에서 유사 서브시퀸스 검색을 위해 사용되었던 타임 워핑 변환 기법(time-warping transformation)을 변형란 알고리즘이다. 또한 제안하는 알고리즘은 움직임 궤적을 모델링하기 위해 사용되는 단일 속성(property)인 각도뿐만 아니라, 거리와 시간과 같은 다중 속성을 지원하며, 사용자 질의에 대해 유사 부분 움직임 궤적 검색을 가능하게 하는 근사 매칭(approximate matching)을 지원한다

• Proceedings of the Korean Information Science Society Conference
• /
• 2001.10a
• /
• pp.178-180
• /
• 2001

모양 기반 검색은 주어진 질의 시퀸스의 요소 값에 상관없이, 모양이 유사한 시퀸스 혹은 부분시퀸스를 찾는 연산이다. 본 논문에서는 시프트, 스케일링, 타임 워핑 등 동일 모양 변환의 다양한 조합을 지원할 수 있는 새로운 모양 기반유사 검색 모델을 제안하고, 효과적인 유사 부분 시퀸스 검색을 위한 인덱싱과 질의 처리 방법을 제안한다. 또한 실세계의 증권데이터를 이용한 다양한 실험 결과에 의하여, 본 방식이 질의 시퀸스와 유사한 모양의 모든 서브시퀸스를 성공적으로 찾는 것은 물론 순차검색 방법과 비교하여 매우 빠른 검색 효율을 가짐을 보인다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2011.04a
• /
• pp.1226-1228
• /
• 2011

본 논문에서는 유사 시퀀스 매칭에서 미리 버림 계산의 효율적인 방법을 제안한다. 미리 버림은 유사 시퀀스 매칭에서 유클리디안 거리 계산 도중 거리 계산 값이 허용치보다 큰 경우 나머지 거리 계산을 하지 않는 방법이다. 기존의 방법은 시퀀스 첫 엔트리를 시작으로 하여 유클리디안 거리 계산을 진행한다. 이 방법은 데이터 고려 없이 계산이 진행되기 때문에 데이터의 특성에 따라 효과가 크게 다른 점을 보인다. 본 논문에서는 미리 버림의 효과를 최대화 시키기 위해 유클리디안 거리 계산 시작점을 오프셋이라 정의하고, 이를 데이터 특성에 맞게 조절하는 방법을 제안한다. 실험 결과, 제안한 오프셋 조절 미리 버림 방법이 대용량의 데이터 베이스 기반 시스템에서 기존 기법에 비해 좋은 성능 향상시킨 것으로 나타났다.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.10 no.1
• /
• pp.75-80
• /
• 2010

Noise handing is very important in voice recognition of vehicle environment. that has been studying about to hardware and software approach. hardware method that is noise filter circuit design, basically using Low-pass filter. it was shown a good result. and the side of software that has been developing about to algorithm for Noise canceler, NN(neural network), etc. in this paper we have analysis about to classified parameter pattern matting level for voice recognition on car noise environment that use of DTW(Dynamic Time Warping) which is applicable time series pattern recognition algorithm.

• Journal of KIISE:Computing Practices and Letters
• /
• v.16 no.10
• /
• pp.949-954
• /
• 2010

Computing the rotation-invariant distance between image time-series is a time-consuming process that incurs a lot of Euclidean distances for all possible rotations. In this paper we propose an innovative solution that significantly reduces the number of Euclidean distances using the triangular inequality. To this end, we first present the notion of self rotation distance and show that, by using the self rotation distance with the triangular inequality, we can prune many unnecessary distance computations. We next present that only one self-rotation is enough for all self-rotation distances required. Experimental results show that our self rotation distance-based methods outperform the existing methods by up to an order of magnitude.

• The KIPS Transactions:PartB
• /
• v.10B no.3
• /
• pp.257-264
• /
• 2003

In the context of a dynamic trading environment, the ultimate goal of the financial forecasting system is to optimize a specific trading objective. This paper proposes a two-phase (extraction and filtering) stock trading system that aims at maximizing the rates of returns. Extraction of stocks is performed by searching specific time-series patterns described by a combination of values of technical indicators. In the filtering phase, several rules are applied to the extracted sets of stocks to select stocks to be actually traded. The filtering rules are automatically induced from past data. From a large database of daily stock prices, the values of technical indicators are calculated. They are used to make the extraction patterns, and the distributions of the discretization intervals of the values are calculated for both positive and negative data sets. We assumed that the values in the intervals of distinctive distribution may contribute to the prediction of future trend of stocks, so the rules for filtering stocks are automatically induced from the data in those intervals. We show the rates of returns when using our trading system outperform the market average. These results mean rule induction method using distributional differences is useful.

• Journal of KIISE:Databases
• /
• v.28 no.1
• /
• pp.15-30
• /
• 2001

서브시퀀스 매칭은 질의 시퀀스와 유사한 서브시퀀스를 가지는 데이터 시퀀스와 해당 서브시퀀스의 위치를 찾는 문제이다. 본 논문에서는 윈도우를 구성하는 방법의 이원성을 이용한 새로운 서부시퀀스 매칭 방법인 Dual-Match는 윈도우를 구성하는 방법에 있어서 Faloutsos 등이 사용한 방법(간단히 FRM 이라한다)의 이원적 접근법이다. 즉, FRM에서는 데이터 시퀀스를 슬라이딩 윈도우로 나누고 질의 시퀀스를 디스조인트 윈도우로 나누는 방법을 사용한 반면, Dual-Match에서는 데이터 시퀀스를 디스조이트 윈도우로 나누고 질의 시퀀스를 슬라이딩 윈도우로 나누는 방법을 사용한다. FRM은 색인에 필요한 저장공간을 줄이기 위하여 개별 점 대신 최소 포함 사각형만을 저장함으로 인하여 많은 착오해답(유사하지 않은 후보 서브시퀀스)을 발생시켰다. Dual-Match는 FRM과 비슷한 크기의 저장공간에 개별 점을 직접 저장함으로써 이 문제를 해결한다. 실험결과, Dual-Match는 많은 경우에 있어서 FRM에 비하여 후보 개수를 크게 줄이고 성능을 향상시켰다. 특히, 선택률이 낮은 경우($10^{-4}$이하)에는 후보 개수를 최대 8800배 까지 줄이고, 페이지 액세스 횟수를 최대 26.9배까지 줄였으며, 성능을 최대 430배까지 향상시켰다. 또한, 동일한 크기의 색인을 생성하는데 있어서 Dual-Match는 FRM보다 4.10~25.6배 빠르게 색인을 구성하였다. 이는 색인 구성시에 CPU 오버헤드의 많은 부분을 차지하는 저차원 변환의 횟수를 FRM에 비해 크게 줄이기 때문이다. 이 같은 결과로 볼 때, Dual-Match는 대용량 데이터베이스에 대한 서부시퀀스 매칭의 성능을 크게 향상시킬 수 있는 획기적인 연구 결과라 믿는다.