A new pattern matching algorithm named Phase Only Correlation (POC) is proposed. POC is defined as modified correlation such that the magnitude components of the Fourier transformed images are replaced with a constant. Discrimination capability of POC is far superior to that of correlation in that the autocorrelation of POC becomes Dirac's delta function. POC is easily integrated into an LSI, and will be popular in various application areas because of its superb characteristics and wide range of applicability.

View full abstract

We have proposed a facial identification system based on isodensity maps and human-face extraction using modified HSV color system, and the effectiveness of this method has been confirmed. In the conventional system, however, if the object whose color is the same as human skin exists in the background, the extraction of human face has sometimes filed. In this paper, an algorithm by which the candidate areas of persons are merged using hue and Quasi-chroma information is described. Experimental results of personal identification system using isodensity maps are also described. From the experimental results, an 95.0% extraction accuracy for 40 pictures has been obtained. And a 99.9% identification accuracy for 110 same person pairs and 11990 different person pairs has been obtained.

View full abstract

This paper presents the PASSFACE system which is a lock-control system using face identification. If one can use his (her) face as a key, it will be very convenient to open a lock. The PASSFACE system identify a face image for judgment whether it should unlock a door or not. The face identification consists of two main processes ; normalization of the size and the inclination of the face image and matching of characteristic face components (eye, nose and mouth).

View full abstract

Speaker recognition is the process of automatically recognizing who is speaking. This paper introduces basic speaker recognition methods and explains a text-prompted speaker verification method. The text-prompted method, in which the system prompts each user with a new key sentence every time the system is used, is widely used these days because it is unnecessary to worry about the system being fooled by recordings of key words or sentences uttered by the user. Moreover, this paper shows that recent speaker recognition techniques achieve high performance with speech data recorded using a high-quality microphone in a noise-free room, but still have some technical problems related to background noise, line distortion, and a priori threshold. It also introduces some technical issues that remain to be solved in the future.

View full abstract

A method of writer verification using weighted feature parameters is proposed. In the proposed method, feature parameters extracted from handwriting information are classified into several categories and each category is weighted properly based on the importance for individual verification. In case that a feature parameter is characteristic and informative, a large amount of weighting is given. In this paper, the reliability of the proposed method is also discussed with test results using Japanese writing data.

View full abstract

This paper proposes a method to easily express a schooling behavior of fishes which is one of the basic defence behaviors. The method obtains the schooling behavior parameters which are suitable for the virtual world by teaching the fish that it will not be attacked when there are in a school. As a result, we confirmed that the fishes which are swimming independently in the begnning, gradually form into schlool. Also, the method of expression chain behavior which is one of the features of schooling behavior is proposed. In this method, the velocity change of the nearest fish acts, as a signal causing other fishs to evade even though predetor is not perceived. As a result, we show that the avoidance behavior of fishes whcih perceives predator causes the school to disperse radially.

View full abstract

This paper describes an LSI developed for the DES or TripleDES encryption algorithm. It handles the encryption of file data stored in memory, encryption of data used in audio-visual services conforming to ITU-T Recommendation H. 233,and real-time encryption of voice data, for example, for digital telephones. It also provides a security feature that prevents unauthorized reading and alteration of the encryption key. It has achieves a maximum throughput of 12.5 Mbytes/sec when encoding file data stored in memory.

View full abstract

The geometry modeling in computer graphics has been very difficult and have needed a lot of work as usual. The facial geometry has particular compleity and personality. In this paper, a generic model is used for the facial modeling method and is deformed and fitted along facial feature points to make the personal model. This paper also describes the algorithm to extract facial feature points from frontal view image. This algorithm is composed of the region segmentation techniques using color information to select eye/eyebrow/lip regions, and the heuristical extraction of facial feature points within local regions. And the experiment results using actual face image shows the error to the manual modeling is slight.

View full abstract

A considerable number of studies have been made on the facial expression recognition techniques in pshycological field, engineering and so on. However, it remains an unsettled problem that trade-off between recognition accuracy and calculation cost. It disturbs realizing real-time processing. In the last few years, real-time recognition systems which use a high-speed graphics workstation or a transputer have been seen. They need such a high-end computer system, so it is difficult to use it as a simple interface between computer and human. In this paper, we propose a real-time facial expression recognition method on the assumption that it runs on generic (low-cost) work-station or PC with video capture function. A face extraction is based on simple temporal differential image. One-dimensional correlation matching method is used for a feature tracking. And performing discrete cosine transform (DCT) to image, calculated coefficients in term of festure vectors are given to the neural network. In the user depended expression classification experiments, we confirm that the accuracy of our method is above 90% for five expression categories.

View full abstract

In visual communication systems in which 3D images of humans and scenes need to be reproduced so that the participants can feel as if they share one common space, fidelity of reproduction of facial expressions is very important. In this paper, we propose a new expression reproduction method based on artistic anatomy. Artists create reference facial expression 3D models and generate other 3D expressions by mixing the reference expressions. Then before telecommunications, the parameters representing the mixtures are related to the facial expression parameters to be obtained by the expression detection system so that detected expressions are reproduced by using the relationship. Promising experimental results are show.

View full abstract

Iris recognition is the personal identification which used the iris of a human eye. The study of iris recognition began at the beginning of the 1980's in U.S.A.. It is a new technology which has realized by recent digital image processing. We describe the structure of iris, merits of iris recognition and mechanism of recognition. In addition, the equipment which we are developing at present is introduced. This equipment is able to automatically capture an iris from a distance about 1m and to recognize it.

View full abstract