An accurate automatic personal identification is critical to a wide range of application domains such as access control, electronic commerce, and welfare benefits disbursement. Traditional personal identification methods (e.g., passwords, and PIN) suffer from a number of drawbacks and are unable to positively identify a person. Biometrics refers to automatic identification of an individual based on her distinct physiological and/or behavioral traits. While biometrics is not an identification panacea, it is beginning to provide very powerful tools for a variety of new applications (e.g., cellular phones, smart cards and international border control) requiring positive identification. This paper attempts to summarize important research issues in biometrics.

Personal authentication technologies will be necessary to ensure security of electronic transactions over open networks. Although biometric authentication is one of the most efficient approaches, accuracy of the biometric authentication is affected by the environment of the data acquisition, procedural parameters, and so on. There is as yet no means of giving a fair comparison of the accuracy of products of biometric vendors. Therefore, a standardization of the accuracy evaluation is necessary. This paper gives a standardization of the accuracy evaluation for the fingerprint verification system.

This paper discusses an application of fingerprint identification technology to enhanced human-machine interaction, and also to information systems, specifically to a mobile authentication terminal for secure networked services and to digital appliances. A "Fingerprint User Interface (FpUI)," exploits information regarding not only who put a finger on its sensor but which specific finger it was. With this user-friendly interface, a user can assign commands, data objects, status, or personalized settings to individual fingers. A functional architecture for a mobile authentication terminal, "Pocket-PID," with fingerprint identification capability is proposed which features an easy-to-use FpUI and high security, where the identification function is totally enclosed within the unit. This enables a user's identity authenticated without any possibility of actual fingerprint data being disclosed. The Pocket-PID facilitates implementation of networked services based on secure biometric user identification.

Smart cards and biometrics can be effectively combined for personal authentication over an open network. The combination is achieved as two-step authentication in which the smart card is authenticated based on a public key infrastructure, and the card holder is authenticated using the template stored in the smart card based on the biometric data. The biometric verification has to be executed in the card for security purposes. This paper describes a fingerprint verification method based on a popular biometric verification technique that can be embedded in a smart card. The prototype system that uses this verification method can verify fingerprints in a few seconds by using the data stored on the smart card.

Recently, biometrics such as a person's fingerprint, face, and voice has come to be used for personal authentication. At present, most biometrics authentication systems depend on verification (one-to-one matching) because such verification takes a short period of time and is expected to provide a quick response. In these systems, however, every single user has to enter an ID number for each authentication session and might feel incovenienced as a result. To improve the operation efficiency, identification (one-to-many matching) is required, but identification is currently assumed to require much more time than verification (i.e., the response time is not practical). After probing these problems, we developed a new method to achieve identification in a short period of time. This method shortens the response time by using a matching score matrix, which is constructed in the enrollment phase. The proposed method is shown to need only about 45 one-to-one matchings to identify data in a database with two thousand fingerprints, a count much less than by conventional methods.

Fingerprint recognition is a well-researched problem, and there are several highly accurate systems commercially available. However, this biometric technology still suffers from problems with the handling of bad quality prints. Recent research has begun to tackle the problems of poor quality data. This paper takes a new approach to one problem besetting fingerprints--that of distortion. Previous attempts have been made to ensure that acquired prints are not distorted, but the novel approach presented here corrects distortions in fingerprints that have already been acquired. This correction is a completely automatic and unsupervised operation. The distortion modelling and correction are explained, and results are presented demonstrating significant improvements in matching accuracy through the application of the technique.

Personal identity verification has a great variety of applications including access to computer terminals, buildings, credit card verification as well as EC. Algorithms for personal identity verification can be roughly classified into four categories depending on static/dynamic and biometric/physical or knowledge based. Finger prints, iris, retina, DNA, face, blood vessels, for instance, are static and biometric. Algorithms which are biometric and dynamic include lip movements, body movements and on-line signatures. Schemes which use passwords are static and knowledge based, whereas methods using magnetic cards and IC cards are physical. Each scheme naturally has its own advantages and disadvantages. A new algorithm is proposed for pen-input on-line signature verification incorporating pen-position, pen-pressure and pen-inclinations trajectories. A preliminary experiment is performed on a data base consisting of 293 genuine writings and 540 forgery writings, from 8 individuals. Average correct verification rate was 97.6% whereas average forgery refection rate was 98.7%. Since no fine tuning was done, this preliminary result looks very promising.

A new approach for extracting significant characteristic within speech signal for distinct speaker is presented. Based on the multiresolution property of wavelet transform, quadrature mirror filters (QMFs) derived by Daubechies is used to decompose the input signal into varied frequency channels. Owning to the uncorrelation property of each resolution derived from QMFs, Linear Predict Coding Cepstrum (LPCC) of lower frequency region and entropy information of higher frequency region for each decomposition process are calculated as the speech feature vectors. In addition, a hard thresholding technique for lower resolution in each decomposition process is also used to remove the effect of noise interference. The experimental result shows that by using this mechanism, not only effectively reduce the effect of noise inference but improve the recognition rate. The proposed feature extraction algorithm is evaluated on MAT telephone speech database for Text-Independent speaker identification using vector quantization (VQ). Some popular existing methods are also evaluated for comparison in this paper. Experimental results show that the performance of the proposed method is more effective and robust than that of the other existing methods. For 80 speakers and 2 seconds utterance, the identification rate is 98.52%. In addition, the performance of our method is very satisfactory even at low SNR.

This paper presents a new approach to modeling speech spectra and pitch for text-independent speaker identification using Gaussian mixture models based on multi-space probability distribution (MSD-GMM). MSD-GMM allows us to model continuous pitch values of voiced frames and discrete symbols for unvoiced frames in a unified framework. Spectral and pitch features are jointly modeled by a two-stream MSD-GMM. We derive maximum likelihood (ML) estimation formulae and minimum classification error (MCE) training procedure for MSD-GMM parameters. The MSD-GMM speaker models are evaluated for text-independent speaker identification tasks. The experimental results show that the MSD-GMM can efficiently model spectral and pitch features of each speaker and outperforms conventional speaker models. The results also demonstrate the utility of the MCE training of the MSD-GMM parameters and the robustness for the inter-session variability.

In this paper we propose an integration of face identification and facial expression recognition. A face is modeled as a graph where the nodes represent facial feature points. This model is used for automatic face and facial feature point detection, and facial feature points tracked by applying flexible feature matching. Face identification is performed by comparing the graphs representing the input face image with individual face models. Facial expression is modeled by finding the relationship between the motion of facial feature points and expression change. Individual and average expression models are generated and then used to identify facial expressions under appropriate categories and the degree of expression changes. The expression model used for facial expression recognition is chosen by the results of face identification.

This paper presents a scale invariant face detection and classification method which uses shift invariant features extracted from a Log-Polar image. Scale changes of a face in an image are represented as shift along the horizontal axis in the Log-Polar image. In order to obtain scale invariant features, shift invariant features are extracted from each row of the Log-Polar image. Autocorrelations, Fourier spectrum, and PARCOR coefficients are used as shift invariant features. These features are then combined with simple classification methods based on Linear Discriminant Analysis to realize scale invariant face detection and classification. The effectiveness of the proposed face detection method is confirmed by experiments using face images captured under different scales, backgrounds, illuminations, and dates. To evaluate the proposed face classification method, we performed experiments using 2,800 face images with 7 scales under 2 different backgrounds and face images of 52 persons.

This paper describes a biometric-based key generation method and its application to a secure communication system. In the proposed method, a personal key which is unique to each user is generated by extracting his/her biometric information. Using the generated personal key, a secure communication system which has the functions of confidentiality and user authentication is realized. As an example of the proposed method, we introduce a personal key generation method based on one's handwriting, and a secure telewriting system which enables the encryption of handwriting information as well as the authentication of a writer. Some simulation results indicate the possibility of realizing the above functions by using a writer's personal key.

This paper proposes the person authentication scheme with the images of his or her face. In the scheme, the CCD camera picks up the images of the person's face casted by the lights from the various directions. The lights can be switched in a random order so fast that you cannot change the forged photographs simultaneously with the switches of the lights. This scheme is then highly secure and has a high cost performance. There are some appropriate applications of the scheme, such as the access control system of entering a room. The security and the efficiency of the scheme is checked by the performance of the false rejection rate and the false acceptance rate.

Emerging multimedia technologies introduce the prevalent multicast transmission, and the multicast tree is determined using the time-invariant network parameters. This paper addresses the time-varying multicast tree problem and presents path selection heuristics for multicast routing to determine an alternative path for real-time applications. A network is partitioned into the optimal region, the disjoint region, and the edge cutset if a branch of the multicast tree meets the un-guaranteed QoS condition. The path selection heuristics operate during the multicast session phase to efficiently select an alternative routing path containing an edge in the edge cutset to connect the multicast tree again. The source-based heuristics PS-SPT finds the path for minimal source-to-destination delay and the sharing-based heuristics PS-DDMC for minimal total cost. These path selection heuristics can efficiently provide solutions to keep the multicast transmission reliable. Simulation results also show that the proposed heuristics can provide effective good solutions for real-time multicast transmission. PS-SPT can select a path with optimal source-to-destination delay and PS-DDMC can select a path with optimal total cost.

In this paper we propose a learning algorithm to enhance the fault tolerance of feedforward neural networks (NNs for short) by manipulating the gradient of sigmoid activation function of the neuron. We assume stuck-at-0 and stuck-at-1 faults of the connection link. For the output layer, we employ the function with the relatively gentle gradient to enhance its fault tolerance. For enhancing the fault tolerance of hidden layer, we steepen the gradient of function after convergence. The experimental results for a character recognition problem show that our NN is superior in fault tolerance, learning cycles and learning time to other NNs trained with the algorithms employing fault injection, forcible weight limit and the calculation of relevance of each weight to the output error. Besides the gradient manipulation incorporated in our algorithm never spoils the generalization ability.

This letter presents a method for reducing power dissipation in a protocol converter. The communication protocol of a VLSI chip hierarchically consists of several sub-protocols and only one of them can be actively working at any given time. In general, protocol converters are implemented by dual protocols of the initially given protocols which are to be interfaced. If the duals of those sub-protocols are implemented in separate modules, we can separate active modules and inactive modules on the fly since only one of the modules can be active at a time. The active/inactive state of a module can be monitored by the control signals that represent the execution of the protocol corresponding to the module. Power reduction can be achieved by dynamically suppressing the clock supply to inactive modules. To trade-off the power reduction rate against the area overhead, the module granularity must be properly chosen. For this purpose, we implement the duals of the atomic protocols in the same module if their state graphs share states except the initial state. Our experimental results show that this method provides significant savings in power consumption of between 18.4% and 92.1% with a 5.3% area overhead.

To improve the unstable performance of the traditional keyword-based search engine due to ambiguities of a natural language such as synonymy and /or polysemy, we have developed a new advanced DLSI (differential latent semantic index) space based probabilistic information retrieval system. The new method exploits a most likelihood posteriori function providing a measure of reliability in retrieving a document in the database having a closest match with another document of a query. Our simple experiment gives a supporting evidence for the validity of the theory, which is capable of capturing the intricate variability in word usage contributing to a more robust context contingent search engine.

This paper proposes a fast algebraic codebook search for DSVD applications. In this method, the codebook search is simplified by reducing the number of possible position combinations using a mean-based track threshold multiplied by heuristically determined optimum threshold factor. And, to guarantee a complexity requirement of DSVD, the maximum number of searching position combinations is limited to 320. The proposed method reduced computational complexity considerably, compared with G.729 with a slight degradation of SNR. Particularly, it shows better speech quality with lower complexity than G.729A.

MPEG-4 emphasizes on coding efficiency and allows for content-based access and transmission of arbitrary shaped object. It addresses the encoding of video object using shape coding, motion estimation, and texture coding for interactivity, high compression ratio, and scalability. In this letter, an advanced object-adaptive vertex-based shape coding method is proposed for encoding the shape of video objects. This method exploits octant-based representation to represent the relation of adjacent vertices and that relation can be used to improve coding efficiency. Simulation results demonstrate that the proposed method may reduce more bits for closely spaced vertices.