Yillbyung Lee

Combination of multiple classifiers for the customer's purchase behavior prediction

In these days, EC companies are eager to learn about their customers using data mining technologies. But the diverse situations of such companies make it difficult to know which is the most effective algorithm for the given problems. Recently, a movement towards combining multiple classifiers has emerged to improve classification results. In this paper, we propose a method for the prediction of the EC customers purchase behavior by combining multiple classifiers based on genetic algorithm. The method was tested and evaluated using Web data from a leading EC company. We also tested the validity of our approach in general classification problems using handwritten numerals. In both cases, our method shows better performance than individual classifiers and other known combining methods we tried.

Iris recognition using collarette boundary localization

There has been a rapid increase in the need of accurate and reliable personal identification infrastructure in recent years, and biometrics has become an important technology for security. The iris recognition system consists of four-process: image acquisition, preprocessing, feature extraction and identification or verification. In this paper, we propose the methods for localizing the iris area between the inner boundary and the collarette boundary, to remove unnecessary areas and to increase the recognition rate. For finding the collarette boundary, histogram equalization and a high pass filter, after using an one-dimensional DFT, are applied to the image. The collarette boundary is found using statistical information from the image, which removes low-frequencies, and, finally, the iris is localized between the inner boundary and the collarette boundary. The iris is localized by two kinds of methods, and the recognition rates were compared. The recognition rate was evaluated by using DWT and SVM. These show that the iris localization by the proposed methods contains more information than the previous methods and improves the recognition rate.

Biometric authentication system using reduced joint feature vector of iris and face

In this paper, we present the biometric authentication system based on the fusion of two user-friendly biometric modalities: Iris and Face. Using one biometric feature can lead to good results, but there is no reliable way to verify the classification. In order to reach robust identification and verification we are combining two different biometric features. we specifically apply 2-D discrete wavelet transform to extract the feature sets of low dimensionality from iris and face. And then to obtain Reduced Joint Feature Vector(RJFV) from these feature sets, Direct Linear Discriminant Analysis (DLDA) is used in our multimodal system. This system can operate in two modes: to identify a particular person or to verify a persons claimed identity. Our results for both cases show that the proposed method leads to a reliable person authentication system.

On the individuality of the iris biometric

Biometric authentication has been considered a model for quantitatively establishing the discriminative power of biometric data. The dichotomy model classifies two biometric samples as coming either from the same person or from two different people. This paper reviews features, distance measures, and classifiers used in iris authentication. For feature extraction we compare simple binary and multi-level 2D wavelet features. For distance measures we examine scalar distances such as Hamming and Euclidean, feature vector and histogram distances. Finally, for the classifiers we compare Bayes decision rule, nearest neighbor, artificial neural network, and support vector machines. Of the eleven different combinations tested, the best one uses multi-level 2D wavelet features, the histogram distance, and a support vector machine classifier.

New focus assessment method for iris recognition systems

In this paper, we propose a new focus assessment method (focus method) for iris recognition systems, which combines the wavelet transform method and the SVM (support vector machine) [Vapnik, V., 1998. Statistical Learning Theory. John Wiley & Sons, NY, USA]. The wavelet-based method estimates focus values by using the ratio of high and low-frequency sub-band averages. The SVM find optimal decision boundary between focused and defocused image with image brightness and focus value as input values. The proposed method has shown distinctive advantages in terms of the following four points. First, by using the proposed wavelet-based method, it detects omni-directional high-frequency which is the characteristic of iris patterns. Second, proposed method can detect slight defocusing of images by using the average of high and low-frequency sub-bands simultaneously. Third, the SVM reduces the error rate of the wavelet-based method by finding the optimum threshold. Fourth, processing time is reduced. It is 213 times faster than the maximum speeds compared with previous spatial domain methods [Kang, B., Park, K., 2006. A study on fast iris restoration based on focus checking. In: LNCS (AMDO 2006), vol. 4069, pp. 19-28; Daugman, J.G., 2004. How iris recognition works. IEEE Trans. Circuits Systems Video Technol. 14, 21-30; Wei, Z., Tan, T., Sun, Z., Cui, J., 2005. Robust and Face Assessment of Iris Image Quality. ICB. pp. 464-471]. For making the reference criteria to compare with the result of focus measures, we propose a schematic model for evaluating the pixel size of SR (specular reflection) in the cornea. To compare with the performance rate, Daugmans kernel, Kangs kernel, Weis kernel and Kautskys method are tested with the proposed method using five different types of data. The proposed method showed the best performance and has shown 3.03% of ER (error rate) using the Yonsei database and 2.62% of ER using the UBIRIS.

Discriminant iris feature and support vector machines for iris recognition

In an iris recognition system, the size of the feature set is normally large. As dimensionality reduction is an important problem in pattern recognition, it is necessary to reduce the dimensionality of the feature space for efficient iris recognition. In this paper. we present one of the major discriminative learning methods, namely, Direct Linear Discriminant Analysis (DLDA). Also, we apply the multiresolution wavelet transform to extract the unique feature from the acquired iris image and to decrease the complexity of computation when using DLDA. The Support Vector Machines (SVM) approach for comparing the similarity between the similar and different irises can be assessed to have the features discriminative power. In the experiments, we have showed that that the proposed method for human iris gave a efficient way of representing iris patterns.

A data reduction method for efficient document skew estimation based on Hough transformation

Document recognition usually requires several preprocessing steps in which skew estimation and correction are critical to get a useful system. This paper proposes an efficient data reduction method to enhance the performance of document skew estimation by using a Hough transformation. The time complexity of the Hough transformation is O(/spl Theta/N), where N is the number of black pixels in a document and /spl Theta/ is the skew estimation range divided by /spl Delta//spl theta/. We might enhance the performance by reducing N or /spl Theta/. The proposed method uses an efficient data reduction method called the modified version of divided horizontal histograms, which reduces the number of black pixels N, while retaining the skewness of document. In order to show the superiority of the proposed method, we have also performed experiments with scanned documents, comparing the result with those of the usual data reduction methods: vertical run-length and connected component methods.

MCML: motion capture markup language for integration of heterogeneous motion capture data

Abstract Motion capture technology is widely used for manufacturing animation since it produces high-quality character motion similar to the actual motion of the human body. However, motion capture has a significant weakness due to the lack of an industry-wide standard for archiving and exchanging motion capture data. It is difficult for animators to reuse and exchange motion capture data with each other. In this paper, we propose a standard format for integrating different motion capture file formats. Our standard format is called Motion Capture Markup Language (MCML). It is a markup language based on eXtensible Markup Language (XML). The purpose of MCML is not only to facilitate the conversion or integration of different formats, but also to allow for greater reusability of motion capture data, through the construction of a motion database storing the MCML documents.

Improved Techniques for an Iris Recognition System with High Performance

We describe in this paper efficient techniques for iris recognition system with high performance from the practical point of view. These techniques range every step for an iris recognition system from the image acquisition step to the final step, the pattern matching, and contain as follows: a method of evaluating the quality of an image in the image acquisition step and excluding it from the subsequent processing if it is not appropriate, a bisection-based Hough transform method on the edge components for detecting the center of the pupil and localizing the iris area from an eye image, an elastic body model for transforming the localized iris area into a simple coordination system, and a compact and efficient feature extraction method which is based on 2D multiresolution wavelet transform. By exploiting these techniques, we can improve the system performance in terms of computationally efficient, and more accurate and robust against noises.

Multiple recognizers system using two-stage combination

Most of the multiple recognizers system use a single combination method, therefore recognition performance depends on the characteristics of selected combination method. In order to solve this dependency problem and to increase recognition performance, we propose a new combination architecture of multiple recognizer that has two combination stages. The proposed system consists of three stages: 1) the recognition stage including 5 recognizers, 2) the first combination stage including 3 combinators which belong to different level, and 3) the combination stage including a simple combinator. We verify the performance of the proposed system using two standard handwritten digit database, CEDAR and CENPARMI, and recognition performance is better than other single combinator systems.