Today, in most of the developed countries the composition of the elderly population is increasing rapidly year by year, and a lot of problems arise because of it. This report is intended to understand the outline of the process of aging and the physiological characteristics related to age. Especially it focuses on the visual functions.

抄録全体を表示

Preferred reproduction and just noticable color difference (JND) in liquid crystal color TV images are studied by the mathod of subject estimation. Flesh, pimento and apple are adapted as color TV images. Number of subjective are 22 (women : l2,men : 10). Average, standerd deviation. dominant wavelength, purity of subjective estimations data are shown and JND is determined from the 75% stgtistic allipses.

抄録全体を表示

Hand-weitten characters can so vary in shape that the recognition rate depends greatly on their quality. To understand bow humans judge their quality. we performed two kinds of experiments : a psychological experiment in which it was suggested that the subjects used common criteria for judging character quality, and an eye-movement experiment in which it was found that the important parts of the character were areas of high stroke density, stroke position and its spacing, and character composition. Based on the results of these experiments. we have proposed new measures for stroke. pixel. line width. aspect ratio, and balance. It was found that the proposed measures were useful factors in determining quality. A system for predicting the quality of hand-written Kanji Charcter was established.

抄録全体を表示

Asynchronous transfer mode (ATM) networks are expected to be widely used as the B-ISDN in the future, and being discussed at CCITT SGXVIII for the standardization. CBR (Continuous Bit Rate) and VBR (Variable Bit Rate) services will be provided on ATM networks. A VBR service is thought to enable transmitted picture quality to be improved. However, cell loss, which may occur in a VBR service, causes catastrophic damages in reproduced image. We have already proposed a layered coding technique in a discrete cosine transform (DCT) coding scheme, in order to accomplish packet loss compensation. Recently, we have constructed the prototype CODEC involved in the layered coding technique. In this paper, we describe about the design of the CODEC and its performance. Especially, we describe the results of the subjective picture quality assessment when cell loss occurrs.

抄録全体を表示

This paper discribes a new video coding method for ATM networks. In this method, video signals are divided into large movement signals and small movement sigals because of large movement signals are more important than small movement signals. So large interframe difference signals are assigned to higher priority packets and small interframe difference signals are assigned to lower priority packets. In consideration on cell loss in ATM networks, the basic structure, predictive precision, refreshing method of frame memories are discussed.

抄録全体を表示

In telecommunication technique, designing a display system that presents us the highly comfortable feeling is the most essential item in all systems. Extended Visual Field with High Resolution Display (EVRD) system has been under development. The linkage between line-of-gaze (LOG) and an active central image is acquired satisfying high stability and sensitivity. Some attemptions to get some clues that enhance the detection algorithm of LOG have been done. Preliminary test was to detect the behavior of eye and head movement in different positions. Second test was to varify the degree of eye motion in differnt directions. Third test was to evaluate several images through the eye motion that relates to objective evaluation. Both behaviours as two freedoms were analyzed by FFT, and then observerd the distribution of these Amplitude Spectrum. The experimental data demonstrates explicitly that how degree the head-eye coordination system is in visually various simations.

抄録全体を表示

A learning procedure for a three layer neural network with a single linear neuron as its output layer is proposed. The network takes (k+1) different stable values, where k is the number of neurons in hidden layer each of which has any non-linear characteristic. The proposed learning procedure consists of two parts, Phase I and Phase II. The former is one for the learning between the hidden and output layers, and the latter is one between the input and the hidden layers. The network is applied to classification of numerals, which shows the effectiveness of the proposed learning procedure.

抄録全体を表示