Catálogo de publicaciones - libros

Pattern recognition and classification problems are most popular issue in machine learning, and it seem that they meet their second golden age with bioinformatics. However, the dataset of bioinformatics has several distinctive characteristics compared to the data set in classical pattern recognition and classification research area. One of the most difficulties using this theory in bioinformatics is that raw data of DNA or protein sequences cannot be directly used as input data for machine learning because every sequence has different length of its own code sequences. Therefore, this paper introduces one of the methods to overcome this difficulty, and also argues that the capability of generalization in this method is very poor as showing simple experiments. Finally, this paper suggests different approach to select the fixed number of effective features by using Support Vector Machine, and noise whitening method. This paper also defines the criteria of this suggested method and shows that this method improves the precision of diagnosing abnormal protein sequences with experiment of classifying ovarian cancer data set.

The paper presents the method of predicting the epileptic seizure on the basis of EEG waveform analysis. The Support Vector Machine and the largest Lyapunov exponent characterization of EEG segments are employed to predict the incoming seizure. The results of numerical experiments will be presented and discussed.

This paper suggests a new method to improve the performance of the pathological/normal voice classification. The effectiveness of the mel frequency-based filter bank energies using the fisher discriminant ratio (FDR) is analyzed. Also, mel frequency cepstrum coefficients (MFCCs) and the feature vectors through the linear discriminant analysis (LDA) transformation of the filter bank energies (FBE) are implemented. In addition, we emphasize the relation between the pathological voice detection and the feature vectors through the FBE-LDA transformation. This paper shows that the FBE LDA-based GMM is a sufficiently distinct method for the pathological/normal voice classification. The proposed method shows better performance than the MFCC-based GMM with noticeable improvement.

In this paper we have introduced new, efficient algorithms for computing one- and two-dimensional Daubechies wavelet transforms of any order, with application to signal processing. These algorithms has been constructed by transforming Daubechies wavelet filters into weighted sum of trivial filters. The theoretical computational complexity of the algorithms has been evaluated and compared to pyramidal and ladder ones. In order to prove the correctness of the theoretical estimation of computational complexity of the algorithms, sample implementations has been supplied. We have proved that the algorithms introduced here are the most robust of all class of Daubechies transforms in terms of computational complexity, especially in two dimensional case.

An efficient branch and bound algorithm for matching protein structures has been developed. The compared protein structures are represented as graphs and a product graph of these graphs is calculated. The resulting product graph is then the input to our algorithm. A maximum clique in the product graph corresponds to the maximum common substructure in the original graphs. Our algorithm, which gives an approximate solution to the maximum clique problem, is compared with exact algorithms commonly used in bioinformatics for protein structural comparisons. The computational results indicate that the new algorithm permits an efficient protein similarity calculation used for protein structure analysis and protein classification.

Hand geometry based biometric verification has proven to be the most suitable and acceptable biometrics trait for medium and low security applications. Hereby a new approach for the personal identification using hand images is presented. Two kinds of biometric indicators are extracted from the low-resolution hand images; (i) palmprint features, which are composed of principal lines, wrinkles, minutiae, delta points, etc., and (ii) hand geometry features which include area/size of palm, length and width of fingers. In the article we focus on feature extraction methods applied to one-sensor multimodal hand-palm biometrics system.

In this paper, we propose a new iris feature watermarking method on face data. This research has following three objectives. First, by using watermarked iris features in addition to face data, the multimodal biometric authentication can be possible, which can increase the authentication accuracy. Second, in case that the saved face data is illegally let out and privacy infringement happens, by checking the inserted iris feature watermark, we can solve the legal responsibility problem about the outflow of face data. In detail, if the iris feature watermark cannot be extracted from the outflow face data, we can insist that the face data is let out from other organization instead of ours. Third, in case that “the iris features need to be transmitted via non-secure and noisy communication channel” [1], it can be invisibly hidden on face data by our method. For the first objective, the face recognition accuracy with iris feature watermark should not be degraded. For the second and third objectives, the inserted iris watermark should be “strong” enough to be extracted irrespective of various kinds of attacks (such as blurring, cropping and rotation attacks) and noise insertion on face data. This research has three advantages compared to previous works. First, to overcome the vulnerability of blurring attack to previous biometric watermarking based on spatial domain, we use the watermarking method in frequency domain. Second, to reduce the degradation of face recognition accuracy due to iris watermarking, we insert the watermark into mid and high frequency bands. Third, through using individual unique iris features for biometric watermarking information and secondary authentication, the security level is much enhanced and we can solve legal responsibility problem about the outflow of face data. Experimental results showed that our algorithm could be used to accomplish above objectives.

Personal identification has lately become a very important issue in the still evolving network society. Biometrics identification methods proved to be very efficient, more natural and easy for users than traditional methods of human identification. Hereby we discuss the idea of human identification based on keystroke dynamics. In the article we focus on our methods of feature extraction from the typing patterns. Moreover, we present satisfactory experimental results and possible applications of keystroke biometrics.

Biometrics-based user authentication has several advantages over traditional password-based systems for standalone authentication applications such as home networks. This is also true for new authentication architectures known as systems, where cryptography and biometrics are merged to achieve high security and user convenience at the same time. Recently, a cryptographic construct, called , has been proposed for crypto-biometric systems. In this paper, we propose an approach to provide both the automatic alignment of fingerprint data and higher security by using a 3D geometric hash table. Based on the experimental results, we confirm that the proposed approach of using the 3D geometric hash table with the idea of the can perform the fingerprint verification securely even with one thousand chaff data included.

Recently, many companies have attempted to adopt biometric technology in their mobile phones. In this paper, we propose a new NIR (Near-Infra-Red) lighting face recognition method for mobile phones by using mega-pixel camera image. This paper presents four advantages and contributions over previous research. First, we propose a new eye detection method for face localization for mobile phones based on corneal specular reflections. To detect these SRs (Specular Reflections) (even for users with glasses), we propose successive On/Off activation of the dual NIR illuminators of mobile phone. Second, because the face image is captured by the NIR illuminator, the nose area can be highly saturated, which can degrade face recognition accuracy. To overcome this problem, we use a simple logarithmic image enhancement method, which is suitable for mobile phones with low processing power. Third, considering the low processing speed of mobile phones, we adopt integer-based PCA (Principal Component Analysis) method for face recognition excluding floating-point operation. Fouth, by comparing the recognition performance using the integer-based PCA to those using LDA (Linear Discriminant Analysis) and ICA (Independent Component Analysis) methods, we could know that the integer-based PCA showed better performance apt for mobile phone with NIR image.