• The KIPS Transactions:PartB
• /
• v.13B no.3 s.106
• /
• pp.309-314
• /
• 2006

Recently, multimedia DBMS is appeared to be the core technology of the information society to store, manage and retrieve multimedia data efficiently. In this paper, we propose a new method for content based-retrieval system using wavelet transform, energy value to extract automatically feature vector from image data, and suggest an effective retrieval technique through this method. Wavelet transform is widely used in image compression and digital signal analysis, and its coefficient values reflect image feature very well. The correlation in wavelet domain between query image data and the stored data in database is used to calculate similarity. In order to assess the image retrieval performance, a set of hundreds images are run. The method using standard derivation and mean value used for feature vector extraction are compared with that of our method based on energy value. For the simulation results, our energy value method was more effective than the one using standard derivation and mean value.

• KIPS Transactions on Software and Data Engineering
• /
• v.9 no.8
• /
• pp.251-258
• /
• 2020

A training dataset for text summarization consists of pairs of a document and its summary. As conventional approaches to building text summarization dataset are human labor intensive, it is not easy to construct large datasets for text summarization. A collection of news articles is one of the most popular resources for text summarization because it is easily accessible, large-scale and high-quality text. From social media news services, we can collect not only headlines and subheads of news articles but also summary descriptions that human editors write about the news articles. Approximately 425,000 pairs of news articles and their summaries are collected from social media. We implemented an automatic extractive summarizer and trained it on the dataset. The performance of the summarizer is compared with unsupervised models. The summarizer achieved better results than unsupervised models in terms of ROUGE score.

• Journal of Internet Computing and Services
• /
• v.12 no.6
• /
• pp.161-170
• /
• 2011

Despite some practical confusion over the variations in the diagram which may be drawn differently depending upon the CASE, the statechart of UML has been widely used to show the dynamic behaviour of the systems. Prior research has employed either simple switch-case statement or the state design pattern to generate source code from the statechart, which may result in varying source codes. This paper made an attempt to formally define the statechart and generate source codes from it. Firstly we cleaned up the XMI which was generated from different CASEs. This XMI has been translated to the EHA to identify automata contained in it. Then the elements of the statechart metamodel were mapped to the java programs. We also verified the quality of source codes by measuring functionality and maintainability. The case employed in this study was the air conditioner. The reason was that the case includes various states and transitions of interest. It was found that XMI was well extracted by removing some legacy codes in the CASE and the source codes were then successfully generated with the concurrency and hierarchy of the statechart. Further research is required to validate it practical significance with a larger case.

• Journal of KIISE:Software and Applications
• /
• v.32 no.11
• /
• pp.1136-1146
• /
• 2005

Fingerprint classification is useful in an automated fingerprint identification system (AFIS) to reduce the matching time by categorizing fingerprints. Based on Henry system that classifies fingerprints into S classes, various techniques such as neural networks and support vector machines (SVMs) have been widely used to classify fingerprints. Especially, SVMs of high classification performance have been actively investigated. Since the SVM is binary classifier, we propose a novel classifier-combination model, multiple decision templates (MuDTs), to classily fingerprints. The method extracts several clusters of different characteristics from samples of a class and constructs a suitable combination model to overcome the restriction of the single model, which may be subject to the ambiguous images. With the experimental results of the proposed on the FingerCodes extracted from NIST Database4 for the five-class and four-class problems, we have achieved a classification accuracy of $90.4\%\;and\;94.9\%\;with\;1.8\%$ rejection, respectively.

• KIISE Transactions on Computing Practices
• /
• v.21 no.3
• /
• pp.257-262
• /
• 2015

As software grows more complex, it contains more bugs that are not recognized by developers. Attackers can then use exploitable bugs to penetrate systems or spread malicious code. As a representative method, attackers manipulated documents or multimedia files in order to make the software engage in unanticipated behavior. Recently, this method has gained frequent use in A.P.T. In this paper, an automatic analysis method to find software security bugs was proposed. This approach aimed at finding security bugs in the software which can arise from input data such as documents or multimedia. Through dynamic taint analysis, how input data propagation to vulnerable code occurred was tracked, and relevant instructions in relation to input data were found. Next, the relevant instructions were translated to a formula and vulnerable input data were found via the formula using an SMT solver. Using this approach, 6 vulnerable codes were found, and data were input to crash applications such as HWP and Gomplayer.

• KIPS Transactions on Software and Data Engineering
• /
• v.3 no.1
• /
• pp.31-36
• /
• 2014

Social network services(SNS) including Twitter and Facebook are good resources to extract various issues like public interest, trend and topic. This paper proposes a method to extract query-related issues by calculating relatedness between terms in Twitter. As a term that frequently appears near query terms should be semantically related to a query, we calculate term relatedness in retrieved documents by summing proximity that is proportional to term frequency and inversely proportional to distance between words. Then terms, relatedness of which is bigger than threshold, are extracted as query-related issues, and our system shows those issues with a connected network. By analyzing single transitions in a connected network, compound words are easily obtained.

• Journal of the Korea Society of Computer and Information
• /
• v.15 no.7
• /
• pp.19-26
• /
• 2010

In this paper, we suggest a efficient handwritten character verification method in on-line environments which automatically analyses two input character string and computes their similarity degrees. The proposed algorithm first applies the circular projection method to input handwritten strings and extracts their representative features including shape, directions, etc. It then calculates the similarity between two character strings by using an improved dynamic time warping (DTW) algorithm. We improved the conventional DTW algorithm efficiently through adopting the branch-and-bound policy to the existing DTW algorithm which is well-known to produce good results in the various optimization problems. The experimental results to verify the performance of the proposed system show that the suggested handwritten character verification method operates more efficiently than the existing DTW and DDTW algorithms in terms of the speed.

• Journal of the Korea Society of Computer and Information
• /
• v.20 no.2
• /
• pp.47-54
• /
• 2015

Emotional reaction of audience can be decided by observing reaction of audience for content. We can use a method to analyze visual data from video camera to detect reaction of audience fast and economically. This paper proposes the method and system to observe audience reaction from visual data of audience and define via neural network. Also we propose a new method to detect automatically an area for audience reaction with face detection to improve a fixed area assignment method which has a limitation not to adapt depending on audiences. Additionally, the evaluation is implemented to show that the proposed method and system is effective. The proposed method showed the performance elevation of 10.5 % (7.75 hit ration) compared to a fixed area assignment method.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.13 no.10
• /
• pp.1997-2003
• /
• 2009

Methods for effective musical score recognition and efficient editing of musical scores are demanded because functions of computers for researches on musical activities become more and more important parts in recent days. In the conventional methods for handling musical scores manually, there are weak points such as incorrect score symbols in input process and requirement of much time to adjust the incorrect symbols. And also there is another weak point that the scores edited by each application program can be remodified by a specific application program only. In this paper, we proposed a method for automatic musical score recognition of printed musical scores in order to make up for the weak points. In the proposed method, staffs in a scanned score image are eliminated by horizontal histogram, noises are removed by 4 directional edge tracking algorithm, and then musical score symbols are extracted by using Grassfire algorithm. The extracted symbols are recognized by hierarchical ART2 algorithm. In order to evaluate the performance of the proposed method, we used 100 musical scores for experiment. In the experiment, we verified that the proposed method using hierarchical ART2 algorithm is efficient.

• Journal of the Korea Society of Computer and Information
• /
• v.26 no.3
• /
• pp.35-42
• /
• 2021

In this paper, we propose a deep learning-based person re-identification method using a three-dimensional RGB-Depth Xtion2 camera considering joint coordinates and dynamic features(velocity, acceleration). The main idea of the proposed identification methodology is to easily extract gait data such as joint coordinates, dynamic features with an RGB-D camera and automatically identify gait patterns through a self-designed one-dimensional convolutional neural network classifier(1D-ConvNet). The accuracy was measured based on the F1 Score, and the influence was measured by comparing the accuracy with the classifier model (JC) that did not consider dynamic characteristics. As a result, our proposed classifier model in the case of considering the dynamic characteristics(JCSpeed) showed about 8% higher F1-Score than JC.