Recently, Distributed Speech Recognition (DSR) systems are widely deployed in Japanese cellular telephone networks. In these systems, personal authentication with voice is strongly desired. In this paper, we present several speaker recognition techniques developed in the University of Tokushima for Distributed Speaker Identification/Verification (DSI/DSV) systems. Especially, we present recent progress on a non-parametric speaker recognition system that is robust to quantization in the distributed systems comparing with conventional speaker recognition systems based on Gaussian Mixture Model (GMM). Evaluation results using the Japanese de facto standard speaker recognition corpus and CCC Speaker Recognition Evaluation 2006 data developed by the Chinese Corpus Consortium (CCC) show higher performance of the proposed method than GMM and VQ-distortion in the European Telecommunications Standards Institute (ETSI) DSR standard environment.

View full abstract

Palmprint based biometric identification has gradually attracted the attention of researchers due to its richness in amount of features. Palmprint contains geometry features, line features, point features, texture features and statistical features. In this paper, simple and effective methodology for palmprint based identification system is proposed. The right hand image is captured using a digital camera without pegging or illumination arrangements. The captured image is aligned using identified key points in the hand and the palmprint region is selected for enhancement and resizing. Different feature extraction methods, namely Discrete Cosine Transform energy features, Wavelet Transform energy features and SobelCode are applied to the resized image to obtain feature vectors. The extracted feature vectors are matched using similarity measurement and feedforward backpropagation neural network. The proposed schemes are tested with hand images from 101 individuals.

View full abstract

Pose estimation is an important part in human face recognition because head rotations will significantly affect the recognition accuracy. However, automated and accurate pose determination still remains an unsolved problem in the research community. In this paper, we propose a novel Elastic Energy Model to tackle this problem, which employs statistical energy contributions of a set of feature points on an input face. It can avoid over-trusting selected anchor points, providing a robust solution to the imprecise feature localization problem that is inevitable in practical applications with cluttered backgrounds. As a general configuration, this model can be easily implemented and extended to other non-rigid objects. Its effectiveness and robustness in automated head pose estimation are investigated in our experiments.

View full abstract

DNA computing often makes use of hybridization, whether for vastly generating the initial candidate answers or amplification via polymerase chain reaction (PCR). The main idea behind DNA computing approaches for solving weighted graph problems is that if the degree of hybridization can be controlled, then it is able to generate more double-stranded DNAs (dsDNAs), which represents the solution to the problem during initial pool generation. Previously, concentration and melting temperature of DNA have been exploited for controlling DNA hybridization during an in vitro computation. In this paper, we present an improved direct-proportional length-based DNA computing (DPLB-DNAC), which combines two characteristics: length and concentration for encoding and at the same time, effectively control the degree of hybridization of DNA. The encoding by length is realized whereby the cost of each path is encoded by the length of the DNA strands in a proportional way. On the other hand, the control of concentration is done by varying the amount of input DNA strands, based on the input graph. The proposed approach shows improvements in terms of materials usage and scalability. The experimental results demonstrate the feasibility of the proposed approach for solving weighted graph problems, such as the shortest path problem.

View full abstract

In this paper, we propose a concavity and convexity enhancement method for noiseless binarization of fingerprint images as a preprocessing step during authentication. Personal authentication with fingerprints is vital and reliable in biometrics based recognition. However, noises could be introduced in fingerprint images capturing, which makes feature extraction and authentication very challenging. Recently, due to the increased need for personal authentication, fingerprint recognition has been widely used in the media. In fingerprint processing, a scanner is usually used to collect the fingerprints. During capture, noise is introduced into the fingerprint image. It is therefore important that preprocessing is carried out, because there are variations in brightness, contrast and noise. To overcome the issues, our proposed method binarizes the fingerprint images. Furthermore, the proposed method produces better results than discriminant analysis or local thresholds methods. Computer Simulations were then carried out to show the effectiveness of the proposed technique. A binarization accuracy of 97.5% was achieved. In authentication, we achieved a perfect False Rejection and False Acceptance Rate of 0.6172%.

View full abstract

In this paper, typing biometrics is applied as an additional security measure to the password-based or Personal Identification Number (PIN)-based systems to authenticate the identity of computer users. In particular, keystroke pressure and latency signals are analyzed using the Fuzzy Min-Max (FMM) neural network for authentication purposes. A special pressure-sensitive keyboard is designed to collect keystroke pressure signals, in addition to the latency signals, from computer users when they type their passwords. Based on the keystroke pressure and latency signals, the FMM network is employed to classify the computer users into two categories, i.e., genuine users or impostors. To assess the effectiveness of the proposed approach, two sets of experiments are conducted, and the results are compared with those from statistical methods and neural network models. The experimental outcomes positively demonstrate the potentials of using typing biometrics and the FMM network to provide an additional security layer for the current password-based or PIN-based methods in authenticating the identity of computer users.

View full abstract

A new fingerprint recognition approach based on features extracted from the wavelet domain is presented. The 64-subband structure proposed by the FBI WSQ standard is used to decompose the frequency of the image. The efficiency of the method is achieved by using the k-nearest neighbor (k-NN) classifier. The result is compared with other image-based methods. For compressed fingerprint images, this proposed method can achieve much lower computational efforts.

View full abstract

A Support Vector Machine (SVM) face identification method using optimized Gabor features is presented in this paper. 200 Gabor features are first selected by a boosting algorithm, which are then combined with SVM to build a two-class based face recognition system. While computation and memory cost of the Gabor feature extraction process has been significantly reduced, our method has achieved the same accuracy as a Gabor feature and LDA based multi-class system.

View full abstract

Based on a coupled circadian oscillators model in mammals, the effects of the structural diversity, the external noise, and the internal noise on synchronization are discussed, respectively. It is found (i) the coupling plays a constructive role in the enhancement of synchronization, while the structural diversity plays a destructive role in the process of synchronous rhythm output. (ii) The synchronization can be enhanced by modulation of the external noise, while it can be suppressed by the internal noise.

View full abstract

Cardiac arrhythmia is mainly due to failure in the coordination among contraction of cardiac muscles. Among several types of arrhythmias, ventricular arrhythmia is being more dangerous and we have chosen this for our research. With the help of mathematical model of ventricle cell, the role of Sodium ion on single cell, 1D array of cells and 2D grid of cells in establishing arrhythmia are studied. In the first study it was observed that the decrease in sodium ion conductance from nominal value in single cell lengthens the action potential duration (APD) and decreases the peak amplitude of AP. In cell pair and 1D array of cells, delay in action potential propagation is observed due to decrease of sodium ion conductance from nominal values. The cells in 60x60 grids are connected discretely via resistive connection, resembles gap junction conductance in real cardiac electrophysiology. In the grid study, normal activation of action potential propagation is first established with an external stimulus (eS1). Along with this eS1 an additional stimulus (eS2) is presented at various locations in order to generate spiral waves, which causes the Re-entrant arrhythmia. Studies are also made to establish spiral waves in the grid by varying sodium conductance.

View full abstract

We have evaluated the narrowband acoustic model based on the extremal sampling technique and the hypothesis, in which no information consisted in the monotonous part from the minimum to the maximum of the bandpass filtered waveform and the vice versa. The sampled extrema were manipulated by the three steps of ±20%, ±40% and ±60% of amplitude errors and more three steps of ±4%, ±9% and ±14% of phase errors. The subjective scores have determined that the margin of error of the extrema exceeded ±20% for amplitude and ±4% for phase. Although the sinusoidal interpolation between two successive extrema of the speech and musical data is better than the other four reconstruction techniques by using the linear function, the sigmoid, the cosine cubed and the cosine to the 0.3^<rd> power, no significant difference is observed for the reproduction of the door noise, suggesting that the noises may be roughly identified in the human auditory system.

View full abstract

In this work, numerical study of pulsating blood flow through the models of the dilated arteries have been studied. A mathematical model of the axially symmetric dilated rigid artery is introduced. The impressible Navier-Stokes equations are solved numerically by using the finite difference method. Effects of the Reynolds number (Re), the Strouhal number (St) and the degree of the dilation (δ) are considered. The calculations based on the geometry and the flow conditions representing a real physiological situation as closely as possible suggest that depending on the values of Re, St and δ. The streamline patterns depict the temporal flow characteristics. Finally, the effect of the wall geometry on the distribution of the wall shear stress is studied. The results show that the dilation disturbs the flow field within an aneurysm and the disturbed blood flow may lead to the formation of a thrombus, increasing the risk of rupture.

View full abstract

Fairness evaluating effects of prescriptions is indispensably importance in disease treatment. The development of Intelligence Technologies nowadays makes it realizable for a moderation of subjective prescription evaluations in medicine. In this paper, we present an application of Self-Organizing Map (SOM) to visualization of herbal prescription effects in Oriental Medicine (OM). The application is used to estimate influence of herbal ingredients prescribed for patients who have several concurrent diseases. Inputs to the SOM are herbal treatment prescriptions, and an output is a map that shows disease states that can be influenced by these prescriptions. Training data is 133 standard prescriptions with 221 popular herbs in 17 disease groups collected from oriental medical text books by experienced doctors. First, the prescription problem in OM is described. Next, the system structure and implementation of the proposed system are shown. Finally, the visualization abilities of the proposed system verified by experienced doctors are presented.

View full abstract

Edema may occur on hands, face, and lower limbs in 20-40% of postpartum women during the early puerperal period. The elevation of lower limbs is the most effective way to treat lower limb edema in postpartum women. New methods have been recently introduced for the care of the lower limb edema, such as footbaths, after-birth exercises, and aromatic oil massage for the lower limbs. However, the effect of massage on reducing lower limb edema is not completely accepted. The objective of this study is to evaluate the change in the lower limb edema after lower limb or dorsolumbar massage with 30-min elevation of the lower limbs. Nineteen postpartum women were divided into two groups, the lower limb massage group (n=9) and the dorsolumbar massage group (n=10). Both groups showed reduction in the lower limb volume as well as the circumference of the lower leg and the big toe, and an increase in the cutaneous blood flow in the tibialis anterior muscle. These results suggest that lower limb or dorsolumbar massage with elevation of the lower limbs reduces lower limb edema.

View full abstract

Since implantable device therapy utilizes microcomputers, soft-errors due to radiations are becoming the dominant reliability failure mechanism. The environmental thermal neutron is one of serious causes of soft-errors. For analyzing thermal neutron fluxes, neutron measurements were conducted in a concrete building of Kyushu University of 11 stories by the use of two types of neutron detectors; one is bare ^3He counter that is sensitive to thermal neutrons, and the other a neutron dose meter that consists of a ^3He counter but sensitive to higher energy neutrons. Cosmic muons, which can produce neutrons via nuclear reactions, were also measured by a parallel plate plastic scintillator counter. The floor variation of flux has been compared with simulation results of the code PHITS and discussed.

View full abstract

Genetic Algorithms (GAs) are well known as evolutionary computation with techniques inspired by evolutionary biology such as inheritance, mutation, selection, and crossover. But GAs have several problems that is the premature convergence and falling in the local solution. For these problems, we have already proposed Fuzzy Adaptive Search Method for GA (FASGA) that is able to tune the genetic parameters according to the search stage by the fuzzy reasoning. And as the improve method of FASGA, we have also proposed Fuzzy Adaptive Search method for Parallel Genetic Algorithm (FASPGA) and proved that it could avoid premature convergence problem well. However, there are some cases when it is not enough accuracy to describe the stage of evolution, because only the best fitness and average fitness were adopted as inputs of fuzzy rules. Therefore, as an improved method for FASPGA, we propose Evolution History based on Fuzzy Adaptive Search method for Parallel GA (EH-FASPGA). In EH-FASPGA, the evolution degree of each island can be grasped by the evaluation of evolution history. Then genetic parameters are tuned according to the evolution degree of each island.

View full abstract

We examine how a toroidal sequence of the six fundamental colors affects human color impression. In order to investigate the different effects of spatial color sequences, we consider a hexagonal diagram that is a projection of RGB color space. The hexagonal diagram corresponds roughly to the hue circle indicated by both hue and saturation. The toroidal sequence is similar to the hue circle. The projected route area indicates the magnitude of naturalness (as in rainbows) for color sequences. The minimum sequence is similar to the order of colors in rainbows, whereas the non-minimum sequence is completely different. Therefore, we propose a simple fuzzy model of human color impression using the projected route area assisted by route complexity. The complexity is defined by the ratio of the square of envelope route distance to route area. The relationship between route complexity and the number of subjects for projected route area is investigated. We clarified that the majority (approximately over 26%) of subjects of nearly all ages have natural impressions when the minimum route area is large, and propose a simple fuzzy model of the human color impression. This model provides the natural (or unnatural) order of spatial color sequences of several colors.

View full abstract