Jaihie Kim

Age estimation using a hierarchical classifier based on global and local facial features

The research related to age estimation using face images has become increasingly important, due to the fact it has a variety of potentially useful applications. An age estimation system is generally composed of aging feature extraction and feature classification; both of which are important in order to improve the performance. For the aging feature extraction, the hybrid features, which are a combination of global and local features, have received a great deal of attention, because this method can compensate for defects found in individual global and local features. As for feature classification, the hierarchical classifier, which is composed of an age group classification (e.g. the class of less than 20 years old, the class of 20-39 years old, etc.) and a detailed age estimation (e.g. 17, 23 years old, etc.), provide a much better performance than other methods. However, both the hybrid features and hierarchical classifier methods have only been studied independently and no research combining them has yet been conducted in the previous works. Consequently, we propose a new age estimation method using a hierarchical classifier method based on both global and local facial features. Our research is novel in the following three ways, compared to the previous works. Firstly, age estimation accuracy is greatly improved through a combination of the proposed hybrid features and the hierarchical classifier. Secondly, new local feature extraction methods are proposed in order to improve the performance of the hybrid features. The wrinkle feature is extracted using a set of region specific Gabor filters, each of which is designed based on the regional direction of the wrinkles, and the skin feature is extracted using a local binary pattern (LBP), capable of extracting the detailed textures of skin. Thirdly, the improved hierarchical classifier is based on a support vector machine (SVM) and a support vector regression (SVR). To reduce the error propagation of the hierarchical classifier, each age group classifier is designed so that the age range to be estimated is overlapped by consideration of false acceptance error (FAE) and false rejection error (FRE) of each classifier. The experimental results showed that the performance of the proposed method was superior to that of the previous methods when using the BERC, PAL and FG-Net aging databases.

Iris feature extraction using independent component analysis

In this paper, we propose a new feature extraction algorithm based on Independent Component Analysis (ICA) for iris recognition. A conventional method based on Gabor wavelets should select the parameters (e.g., spatial location, orientation, and frequency) for fixed bases. We apply ICA to generating optimal basis vectors for the problem of extracting efficient feature vectors which represent iris signals. The basis vectors learned by ICA are localized in both space and frequency like Gabor wavelets. The coefficients of the ICA expansion are used as feature vector. Then, each iris feature vector is encoded into an iris code. Experimental results show that our proposed method has a similar Equal Error Rate (EER) to a conventional method based on Gabor wavelets and two advantages: first, the size of an iris code and the processing time of the feature extraction are significantly reduced; and second, it is possible to estimate the linear transform for feature extraction from the iris signals themselves.

Cancelable fingerprint templates using minutiae-based bit-strings

It has become critical to protect biometric templates in the current biometric community. One way for doing this is using a cancelable biometric method, which transforms original biometric templates in a non-invertible way and uses those transformed templates to verify a persons identity. In this paper, we propose a new method to generate cancelable bit-strings (templates) from fingerprint minutiae. Our method is to provide a simple mean to generate cancelable templates without requiring for pre-alignment of fingerprints. The main idea is to map the minutiae into a predefined 3 dimensional array which consist of small cells and find out which cells include minutiae. To do this, we choose one of minutiae as a reference minutia and other minutiae are translated and rotated in order to map the minutiae into the cells based on the position and orientation of the reference minutia. After mapping, we set the cells in the 3D array to 1 if they include more than one minutia otherwise the cells are set to 0. A 1D bit-string is generated by sequentially visiting the cells in the 3D array. The order of the 1D bit-string is permuted according to the type of reference minutiae and users PIN so that we can regenerate new templates when we need them. Finally, cancelable bit-strings are generated by changing the reference minutia into another minutia in turn. In the experiments, we evaluate our method using the FVC2004 database and show that the performance is better than that of a previous method.

Fake iris detection by using purkinje image

Fake iris detection is to detect and defeat a fake (forgery) iris image input. To solve the problems of previous researches on fake iris detection, we propose the new method of detecting fake iris attack based on the Purkinje image. Especially, we calculated the theoretical positions and distances between the Purkinje images based on the human eye model and the performance of fake detection algorithm could be much enhanced by such information. Experimental results showed that the FAR (False Acceptance Rate for accepting fake iris as live one) was 0.33% and FRR(False Rejection Rate of rejecting live iris as fake one) was 0.33%.

Biometric key binding: fuzzy vault based on iris images

Recently, crypto-biometric systems have been studied for solving the key management problem of cryptographic systems and protecting templates in biometric systems at the same time. The fuzzy vault system is a well-known crypto-biometric system. We propose a new method of applying iris data to the fuzzy vault. Our research has following two advantages and contributions. First, in order to solve the variation problem of the extracted iris features, we introduce a pattern clustering method. Second, in order to produce unordered sets for fuzzy vault, we use the iris feature extraction algorithm based on ICA (Independent Component Analysis). Experimental results showed that 128-bit cryptographic keys as well as the iris templates were secure with the fuzzy vault scheme.

A novel method to extract features for iris recognition system

In general, the iris recognition systems have used the wavelet transform as feature extraction techniques. Since the wavelet transform does not have the shift-invariant property, the iris features are inconsistently extracted due to the eye image rotation and the inexact iris localization. In this paper, a novel method to extract features is proposed for iris recognition system. Two types of features are obtained from the discrete wavelet frame decomposition. The first one is the global feature which is insensitive to the iris image deformation. The second one is the local feature which can represent the iris local texture. If the global distance between the test image and the stored one in the database is smaller than the threshold value, it is added to the candidates. And then, local matching is performed by Hamming distance. Experimental results show the proposed system could be used for the personal recognition efficiently.

A robust fingerprint matching algorithm using local alignment

This paper describes a minutiae-based fingerprint matching algorithm. Generally, a fingerprint image is nonlinearly deformed by torsion and traction when a finger is pressed on the sensor. This nonlinear deformation changes both position and orientation of minutiae and decreases the reliability of minutiae. Therefore, in matching algorithm using one reference minutiae pair, the reliability of a minutia decreases as the distance from the minutia to the minutia used for alignment increases. The proposed algorithm overcomes this problem by normalizing the distance between minutiae and using local alignment. Experimental results show that the performance of the proposed algorithm is superior to that of using one reference minutiae pair.

Parking Slot Markings Recognition for Automatic Parking Assist System

This paper describes a monocular vision based parking-slot-markings recognition algorithm, which is used to automate the target position selection of automatic parking assist system. Peak-pair detection and clustering in Hough space recognize marking lines. Specially, one-dimensional filter in Hough space is designed to utilize a priori knowledge about the characteristics of marking lines in birds eye view edge image. Modified distance between point and line-segment is used to distinguish guideline from recognized marking line-segments. Once the guideline is successfully recognized, T-shape template matching easily recognizes dividing marking line-segments. Experiments show that proposed algorithm successfully recognizes parking slots even when adjacent vehicles occlude parking-slot-markings severely

A New Method for Generating an Invariant Iris Private Key Based on the Fuzzy Vault System

Cryptographic systems have been widely used in many information security applications. One main challenge that these systems have faced has been how to protect private keys from attackers. Recently, biometric cryptosystems have been introduced as a reliable way of concealing private keys by using biometric data. A fuzzy vault refers to a biometric cryptosystem that can be used to effectively protect private keys and to release them only when legitimate users enter their biometric data. In biometric systems, a critical problem is storing biometric templates in a database. However, fuzzy vault systems do not need to directly store these templates since they are combined with private keys by using cryptography. Previous fuzzy vault systems were designed by using fingerprint, face, and so on. However, there has been no attempt to implement a fuzzy vault system that used an iris. In biometric applications, it is widely known that an iris can discriminate between persons better than other biometric modalities. In this paper, we propose a reliable fuzzy vault system based on local iris features. We extracted multiple iris features from multiple local regions in a given iris image, and the exact values of the unordered set were then produced using the clustering method. To align the iris templates with the new input iris data, a shift-matching technique was applied. Experimental results showed that 128-bit private keys were securely and robustly generated by using any given iris data without requiring prealignment.

Biometric scores fusion based on total error rate minimization

This paper addresses the biometric scores fusion problem from the error rate minimization point of view. Comparing to the conventional approach which treats fusion classifier design and performance evaluation as a two-stage process, this work directly optimizes the target performance with respect to fusion classifier design. Based on a smooth approximation to the total error rate of identity verification, a deterministic solution is proposed to solve the fusion optimization problem. The proposed method is applied to a face and iris verification fusion problem addressing the demand for high security in the modern networked society. Our empirical evaluations show promising potential in terms of decision accuracy and computing efficiency.