Hye-Wuk Jung

Fingerprint classification using the stochastic approach of ridge direction information

Large scale, automatic fingerprint identification systems (AFISs) perform fingerprint classification to improve matching accuracy and reduce the matching time before fingerprint matching. Fingerprints are classified into several classes such as arch (A), whorl (W), left loop (L) and right loop (L). The existing systems generally classify fingerprints based on the information of singular points. This approach is well suited for fingerprints acquired using paper and ink. However, it is not as efficient with recent automatic fingerprint systems because it cannot guarantee that singular points are well extracted since the recent systems have various sized sensors and use multifarious fingerprint acquisition methods. In this paper, a novel approach is proposed to use the fingerprint ridge direction, which is one of the global features. It is a probabilistic approach based on the fingerprint ridge characteristics of each class. FVC2000 DB1 and FVC2002 DB1 databases were used to evaluate the performance of our classification. Furthermore, the effectiveness of applying the probabilistic model to the classification of various exceptional fingerprint patterns was verified.

Topic word selection for blogs by topic richness using web search result clustering

Blogs are one of popular services to publish and archive posts with personal opinions on the web. The topics of a blog can be used for classification, recommendation, opinion mining, and ranking, etc. In this paper, we propose a method for extracting important topic words from a blog. Our method selects topic words by measuring whether the blog includes rich content on the word. To measure the richness of a blog on candidate topic words, we compare web search results by the candidate words with the content of the blog. We used document clustering and cluster matching in order to compare them.

Noisy and incomplete fingerprint classification using local ridge distribution models

Fingerprint images acquired from live-scan devices may have various noises, such as cuts and smears and be incomplete due to shifted and partial scanning. We propose a novel fingerprint classification method that is able to effectively classify noisy and incomplete fingerprints, which are acquired by live-scan devices. Fingerprint images are divided into blocks of 16×16 pixels and representative directional values of each block are extracted. Based on the representative directional values, the core blocks including the core points are identified by core block Markov models. Then, fingerprints are divided into 4 regions with respect to the core blocks and each region is modeled with the distribution of the ridge directional values in its region. Fingerprint classification is carried out by using the regional local models. If a fingerprint is given, each local model determines the probabilities that the given fingerprint belongs to all the fingerprint classes. The final decision on the classification is made by probabilistic integration of the classification results of local models. Since the proposed method analyzes ridges based on blocks of 16×16 pixels and classifies based on regional local models, it can be robustly applied to noisy and incomplete fingerprint images. A performance evaluation based on the live scanned fingerprint databases FVC 2000, 2002, and 2004 shows a good classification accuracy of 97.4%. A regional local model based fingerprint classification method is proposed.We make regional local models from the probability distributions of ridge directions.A classification accuracy based on the live scanned fingerprint databases is 97.4%.The classification performance is high for low quality and incomplete fingerprints.

Personalized expert-based recommender system: training C-SVM for personalized expert identification

In order to improve the performance of the existing recommendation algorithms, previous researches on expert-based recommender systems have exploited the knowledge of experts. However, the previous expert-based recommender systems are limited in that the same experts are suggested for all users. In this paper, we study personalized expert identification problem, assuming each user needs different kinds and levels of expert help. We demonstrate the feasibility of personalized expert-based recommendation; we present and analyze an SVM framework for finding personalized experts.

An Auto Playlist Generation System with One Seed Song

The rise of music resources has led to a parallel rise in the need to manage thousands of songs on user devices. So users have a tendency to build playlist for manage songs. However the manual selection of songs for creating playlist is a troublesome work. This paper proposes an auto playlist generation system considering user context of use and preferences. This system has two separated systems; 1) the mood and emotion classification system and 2) the music recommendation system. Firstly, users need to choose just one seed song for reflecting their context of use. Then system recommends candidate song list before the current song ends in order to fill up user playlist. User also can remove unsatisfied songs from the recommended song list to adapt the user preference model on the system for the next song list. The generated playlists show well defined mood and emotion of music and provide songs that the preference of the current user is reflected.

Behavior Pattern Modeling based Game Bot detection

Korean Game industry, especially MMORPG(Massively Multiplayer Online Game) has been rapidly expanding in these days. But As game industry is growing, lots of online game security incidents have also been increasing and getting prevailing. One of the most critical security incidents is `Game Bots`, which are programs to play MMORPG instead of human players. If player let the game bots play for them, they can get a lot of benefic game elements (experience points, items, etc.) without any effort, and it is considered unfair to other players. Plenty of game companies try to prevent bots, but it does not work well. In this paper, we propose a behavior pattern model for detecting bots. We analyzed behaviors of human players as well as bots and identified six game features to build the model to differentiate game bots from human players. Based on these features, we made a Naive Bayesian classifier to reasoning the game bot or not. To evaluated our method, we used 10 game bot data and 6 human Player data. As a result, we classify Game bot and human player with 88% accuracy.

Document Summarization Using Mutual Recommendation with LSA and Sense Analysis

In this paper, we describe a new summarizing method based on a graph-based and a sense-based analysis. In the graph-based analysis, we convert sentences in a document into word vectors and calculate the similarity between each sentence using LSA. We reflect this similarity of sentences and the rarity scores of words in sentences to define weights of edges in the graph. Meanwhile, in the sense-based analysis, in order to determine the sense of words, subjectivity or objectivity, we built a database which is extended from the golden standards using Wordnet. We calculate the subjectivity of sentences from the sense of words, and select more subjective sentences. Lastly, we combine the results of these two methods. We evaluate the performance of the proposed method using classification games, which are usually used to measure the performances of summarization methods. We compare our method with the MS-Word auto-summarization, and verify the effectiveness of ours.

Richness evaluation of blogs on its topics using a generative model and probabilistic analysis

Nowadays, blogs are one of important web services to publish and share various information. Accordingly, evaluation of various keywords in blogs is one of the important research topics for effective and efficient classification and retrieval of blogs in the blogosphere. In this paper, we propose a method to identify important keywords in a blog. In order to identify such keywords, we consider web context, assuming that the blogs documents are generated from web contexts by proposed generative model. Therefore, if the contexts of keyword on the web are reflected well in the blog, then we may regard the keyword is essential because the blog is rich on the keyword. We clustered the blog articles on the given keyword by several subtopics using LDA (Latent Dirichlet Analysis), and compared the clusters with the web context documents obtained by web search. Finally, we evaluated the richness of blog on each keyword.