This paper describes improvement of face identification performance using synthesized images. Three expression features, smile, anger and scream, were extracted from actual images of facial expression by the eigenspace method. Synthesized face images based on these features were added on training data for personal identification model with support vector machines (SVM). The performance of this model was significantly higher than model trained without expression face images, but significantly lower than model using actual expression images. The results suggested that identification performance also significantly depended on facial expression.
Influence of sound environment during numerical memorizing task is examined by the measures of reaction time (RT) and electroencephalogram (EEG) power spectrum density (PSD). The task consists of memorizing eight 1-digits and then discriminating those from displayed 8-digit. The result suggests that sound environment significantly affects reaction time and theta rhythm's PSD in the pause for rehearsal of memorizing numerals.
In this research, it is examined the difference of the sentence understanding across three display devices; printed material, computer CRT and LCD. As a result, a degree of understanding content for printed material is significantly the highest among three media. LCD shows better performance of the understanding than for CRT. Comparing the performance between Gothic and Mincho font sets, performances for using Gothic font are significantly higher than for using Mincho font across three media. The factor of font size influences the performance in smaller font size for CRT display.
A mobile phone is developing as a personal digital assistants, therefore most current mobile phones can connect to the Internet including Web browsing and email access. This paper describes development and evaluation of communication support system using mobile phone for the distance education by satellite which is provided to many points. All students of the class always can tell their question and comment to the lecturer at the remote room using this system and their own mobile phone. However, it is difficult for a lecturer to recognize those received questions in talking various topics. In this system, all presentation patterns for the lecture were clustered by SOM(self-organizing map), and a structural map was constructed to assist a lecturer in advance. Then each question was displayed for the lecture in a class by superimposing its label to the map.
This system was operated in two-days intensive course as an university's open seminar July 2001. To evaluate this system performance, student's subjective evaluation was measured by questionnaire which consisted of 6 items concerning the mobile phone usage and 15 items concerning this course. All responses for items concerning the mobile phone usage are significantly higher than the neutral without response for "promote to use mobile phone in classroom". This suggests that students recognize the effectiveness of this system despite they do not agree to use mobile phone in classroom. Four factors were extracted from the responses concerning this course by conducting the factor analysis. Significant positive correlation was examined between score of the second factor "understanding the lecture content" and response for "total evaluation of this system".
Those results suggest the evidence of the effectiveness of this system for a distance lecture.
This paper describes the relationship between oculo-motor indices which include pupil size and time ratio for blinking, gazing and saccade. The oculo-motor indices respond to respectively the task difficulty and viewer's eye-movement which is controled by switching frequency of a viewing target. The correlation coefficients between time ratios were depend on the switching frequency.
The purpose of this study is to find out the means of effective expression in Japanese for clear understanding. Four experiments were conducted, which dealt with length of the sentence, connection particle "GA", positioning of the subject, and number of punctuation marks. Subjects were presented sentences and given discrimination tasks. The percentage correct and the reaction time for answer response were measured in all the experiments, and evaluated to examine the effectiveness of the four factors. In the first experiment, length of the sentence was controlled. The reaction times (RTs) for sentences under 50 letters are shorter than that of sentences over 50 letters. RTs for a one-topic sentence are shorter than those for multiple-topic sentences. This suggests that a sentence should be under 50 letters and have one topic. The second experiment, the influence of connection particle "GA" for the sentence comprehension was examined. When two phrases were connected by "GA" used as a conjunction, the percentage correct for the former contents is lower than that for the latter contents. This suggests that "GA" should not be used and a one-topic sentence is desirable. In the third experiment, authors examined the influence of positioning of the subject in the sentence comprehension. When the sentences contain the subordinate clause of the reason, the percentage correct for sentences with the subject just before the predicate is higher than that for sentences with the subject on the beginning of sentences. This suggests that the subject should be placed just before the predicate. The fourth experiment, number of punctuation marks was controlled and measured readability of the sentence comprehension. The degree of comprehension was compared in two groups. One is the sentence without punctuation marks around 31.5 letters. The other is the sentence including them about 11.3 letters. The percentage correct for the former is lower than that for the latter. When a sentence without punctuation marks more than 30 letters, marks should be used.
Pupil size and eye blink reflect mental work-load. This study examine whether eye movement responds user's task difficulty, as same as pupil size and blink can be used as indices of it. The calculation task which required an oral calculation or response, was conducted with ocular following task. Pupil size and blink rate increased with responding to calculation task, both were the higer in correct response[slow] and the highest in incorrect response. Gazing time decreased with blink rate increses. Saccade lengths and saccade occurrece rate decreased under the correct response[slow] and the incorrect response conditions. The results of these experiments provide evidence that eye-movement can be used as an index of user's task difficulty.
This paper describes the change of reading speed for Japanese text according to the switching frequency of a random-dot background. The switching frequencies consist of five levels for shifting in the horizontal or vertical directions. The texts are presented in horizontal and vertical format. The reading speed is the lowest in 30Hz of background switching. Vertical shift is the most significant in reducing the reading speed. It is found that these changes occur on both format.
It is well known that pupil size measuring is influenced by noises and blink. This paper describes a development of estimation model of pupil size for blink artifact. The model development was based on 3 layered perceptron with backpropagation method. First, the model was trained by pupil response to brightness change, and the model could estimate the pupil temporal changes. Second, the model was trained by pupil response with artificial blinks. The model could also estimate the pupillary changes and pupil size while blinks. The model was applied to pupillary changes in viewing TV program. The performance of removing brightness influence was evaluated one-way ANOVA. The sampling rate for pupil size could be improved to 30Hz from 3Hz by using the model processing.
This paper describes development of information retrieval support system for educational materials on the web, and its performance of the retrieval. The feature vectors of the term and the subject and grade were extracted from the teaching guide line for the primary school by singular value decomposition(SVD). The self-organizing maps for the terms and the subject and grade were trained by both feature vectors. The developed information retrieval support system diplays the terms map and the subject map, and terms in high similarity according to a term which is assigned by users. The performance of this system is evaluated by variances of precision score for a subset of educational documents on the web. The variance increase with addition of several terms which show higher similarity to user assigned term. It suggests this system extracts intended documents by assisting terms selection for searching.
Eye-movement and pupillary change has been used as indecies in educational studies and behavioral analysis. Because they reflect the process of human information input, it is possible to use them as indecies for human cognitive load. This paper evaluates cognitive load during driving simulation using eye-movement and pupillary changes. Experimental results based on pupillary size increases, indicate that the cognitive load increases when coupled with task. In addition,saccade rate increased and the time spent on gazing fell from previous cases.
To retrieve and categorize educational information, the relationships between subjects and terms and between two terms were examined by analyzing the frequency of terms that appeared in the teaching guidelines of elementary schools. Using the Singular Value Decomposition method, feature vectors were obtained from the number of occurrences of the terms that appeared in the teaching guidelines. Using these feature vectors, terms and subject areas that contained them were analyzed to indicate the relationships between them. Codebook vectors, that were learned by a Self-Organizing Algorythm with feature vectors, then put the terms and subjects into a 2-dimensional map that described their relationship.
Two textbooks for teaching Japanese as a foreign language were analyzed to extract the relationships between nouns and verbs and the appropriate particles in simple Japanese sentences. A computer based drill to learn Japanese particle usage, that can be used over the internet, was developed by using these relationships. In this drill, the text is displayed with illustrated pictures to indicate the correct verb group and how to correctly use the particle. The average test score for learners with pictures was significantly higher than the score without pictures. The pictures were more effective for those sentences that contained more than one particle. In addition, this method has a significantly larger effectiveness for learners who speak Indo-European languages.
This paper describes the relationship between EEG and pupillary changes in memorizing 8 digits. The independent components were extracted from EEG data by use of ICA(Independent Component Analysis). The contribution of those date to a correct answer was evaluated by a simple discrimination function.
This paper described the correlation between employees' personal data (capabilities obtained before joining the company, performance shown during new employee training and personality as profiled by instructors) at the begining of employment, and the wage ranking (personnel evaluations) in their 10th year of employment. The null hypotheses were rejected at the 5% level of significance between entry records of "Technical expertise," "Mathematics" and "Engilish" socres in the employment examination and the "Basic knowlegde on Information science" course in the employee training and the wage ranking in the 10th year. Multiple regression analysis and Hayashis quantification analysis Type II were cunducted for th classification of wage making form entry and the training records. A 66.3% correct classification level was attained by the latter method of analysis.
Important significant items were "Technical expertise" in the employment examination, "hard Worker" and "Exactness on Work" in their personality profiles, and "Basic knowledge on information science" in new employee training. This shows that basic and fundamental knowledge, including "Technical expertise" acquired before employment and "Basic knowledge on information science" in the new employee training is correlated with the 10th year ratings.
It is possible to evaluate viewer's interest in visual information using pupillary changes caused by mental activity. This paper examines the possibility of estimating viewer's interest and enjoyment of T.V. programs by measuring dynamic pupillary changes, blinking, and subjective interest responses. The primary factors for pupillary changes are derived, for evaluating images of T.V. programs from questionnaires which are based on the SD method. The results showed, in dynamic evaluation, that there is a relationship between pupillary changes and subjective interest response. In evaluating images from T.V. programs, there is a significant difference in pupil size between the higher and the lower group for the factor score of 'Potency', and a difference in blinking between the higher and the lower group for the factor score of 'Activity'. Also there is a significant correlation between subjective interest response and all three factors.
This paper describes the relationship of the nose's temperature change and correctness of the memory task which are figure memory task and word memory task. The mean temperature of the nose for higher correctness group was significantly lower than that for lower correctness group in both memory tasks.
This paper describes a comparative analysis of training effects in four different training styles to study the training methods to enhance training effect. A same training object is used in four different training styles which are (1):schooling using CAI(I), (2):mutual teaching using printed training materials, (3):mutual teaching using CAI(I), (4):self-training t individual local office using CAI(II). Their achievement score values were compared with that of conventional face-to-face lecture style training. The comparison result shows that "(3):mutual training using CAI(I)" has the best training effect among them. Compared with the face-to-face lecture style, it is significant in 1% level either in the group of lower points, int the group of higher points, or in total group. It is also found out that "(4):self-training at individual local office using CAI(II)" is useful in reducing the training time 24%.
Two different delivery systems for distance education by satellite were evaluated by same questionnaire respectively. Two delivery systems are mainly different in transmission system i.e. analog video transmission system and 1.5Mbps digital compressed video transmission system. The overall evaluation in the analog system was significantly higher than that in the digital system. Factor analysis was conducted to extract 6 factors (image quality, presence, Q & A communication, time schedule, content and audio) from 25 questionnaire items. The score of the image quality in the analog system, and the scores of the Q & A communication and the audio in the digital system were significantly higher than those of another system. Multiple regression analysis was also conducted to find out the relationship between overall evaluation and 6 factors. As the result, the image quality and the presence were significantly effective to overall evaluation, but only the presence had a effect in the digital system.
This paper describes the effectiveness of various methods of presentation to include overhead projector(OHP) transparencies, printed materials and verbal instructions. Two courses were used in the experiment, one on Japanese history and the other on world geography. The effects of different types of visual patterns upon the learners' understanding was also analyzed. Sentence patterns were compared with important key word patterns. The results showed that the most effective type of presentation was where the key words were presented using an OHP supplemented by the use of a pointer and verbal instructions. The lowest scores were seen for verbal only presentations.
This paper describes the relationship between children's drawing activities and eye movement. The subjects were 20 third grade elementary school students. They were first asked to draw a picture of a tulip from their imagination. A week later, the same students were asked to draw a tulip again, but this time while looking at a video display of one. Four arts and crafts teachers used a ten-item evaluation sheet to evaluate the resulting drawings. Factor analysis was used to extracted four factors (petals, stem and leaves, position and size, and balance) from each of the teacher's evaluations. A PC was used to measure each students gazing time and frequency, again based upon four categories (petals, stem and leaves, TV display, and canvas). The results showed mean that there is a negative correlation between mean gazing time and teacher evaluation scores. It also showed that shorter gazing time and frequent observation resulted in better drawings. Also, total gazing time and frequency contributed to higher drawing skill performances.
This paper describes an experiment the relationship between word recognition and computer screen text style and location. For the experiment, 16 computer screen positions (4 top to bottom by 4 left to right) were used to show character strings of Hiragana, Katakana or Kanji. The characters were grouped in strings of three to five each, some making words and others being nonsensical. The strings were shown both vertically and horizontally. The subjects' times for locating the previously assigned target screen was measured. The results showed that reactions times were shorter when horizontal strings were shown in the upper left portion of the screen as well as when vertical strings were shown in the upper right portion of the screen. Average reaction times were also shorter for four and five character strings when they were shown vertically rather than horizontally. It was also found that locating words in Hiragana or Katakana is easier than locating words in Kanji. the authors also report on their investigation of eye movement patterns.
If the effects of variations in brightness can be controlled, it is possible to use changes in eye pupil size in the evaluation of educational TV programs. This paper describes an improved, neural-network-based method for studying pupil reactions and the feasibility of evaluating TV programs by investigating changes in pupil size. This neural-network-based reaction model uses M-sequences and Markov-sequences to register changes in pupil size based upon responses to changes in brightness and is shown to be more effective in removing the effects of brightness upon pupil size than have other previously reported methods. The authors also discuss methods for removing the influence of brightness from pupil responses as well as methods for evaluating changes in pupil size. It is also shown that this new method permits the objective evaluation of TV programs through measurement of eye pupil changes during viewing.
Pupil size changes by influence of brightness, but it is known that it also changes when we are interested in something, or when we learn something. Psychological activity and brightness effect on pupil size independently, so psychological activity can be measured if how pupil size changes by influence of brightness is known. In this paper, A linear model and a neural network model is obtained for pupil reaction for brightness. It is shown that these models can remove an influence of brightness change when actual TV programs are presented. The neural network model might correspond to real biological system of eye pupil.
This paper describes the effectiveness of HDTV(high definition TV) display in learning. While twenty geometrical proof problems were shown to 10 subjects, eye movement was measured in understanding each proof. A proof was consisted of a simple line figure and proof text. The subjects were asked to read aloud proof for understanding. The times for understanding each proof were measured. The sum of gazing time to graphics and texts also measured respectively. As the results, in the case of 7 times distance of display height from display, the time to understand the proof displayed on HDTV was significantly shorter than one of NTSC. This effect was found maily in the sum of gazing time to text. The difficulty of problems made the gazing time to texts longer. Those results suggested display resolution had an effect of text readability.
The test scores of students taking regular lectures and those using the PINE-NET System were compared. The PINE-NET system is a nation-wide, communication satellite broadcast systems combined with computer-assisted learning. The results showed that the test scores during the lectures as well as the end of the term were significantly higher for students using the PINE-NET system than for those taking regular classes. This effect was seen across a number of subjects. The relationships between pre- and post-test scores for the two methods could be sorted into four categories. Those subjects which were taught using both methods were also categorized and the effectiveness of the PINE-NET system analyzed. The authors also found a correlation between end-of-semester test scores and entrance examination (covering three subjects) scores. This showed that end-of-semester test scores depend greatly upon the basic abilities of the subjects.
The Effect of stero television presentation on "mental rotation" was examined. When the orientations of two identical or mirror figures are changed and then displayed, mental rotation is the phenomenon that the time taken to decide this difference between the orientations of the two figures. First, the reaction times when presentation were in two and three dimensions were compared, and the effectiveness of the three-dimensional presentation was studied. The result was that, even when displayed in three dimensions, a mental rotation similar that of the two-dimentional case was verified and the reaction time was siginificantly shortened. The result of studying the internal processes showed that three-dimentional presentation was effective from the standpoint of the transformation to the internal representaion. The subjects were divided into two groups based on their reaction times. When the effectiveness of the three-dimensional presetation was studied, three-dimensional presentation was particularly effective for the group with the shorter reaction times. Furthermore, when a part of the presentation figures were marked with color, the speed for rotating the representation became siginificantly faster regardless of the presentation dimension. By studying the subjects, we found colored marking was effective for subjects with long reaction times.
It is well known that pupil size response to both brightness and to mental activity. However, the correlation between these two phenomena is not too clear. this study, therefore is on changes in pupil size in response to mental activity while controlling the effects of brightness. As a first step, pupillary changes were measured at variable brightness with and without verbal instructions designed to stimulate mental activity. No interactions between pupillary changes due to the effects of brightness and those of mental activity were found. As a second step, white gray and black patterns were presented to the subjects and pupil sizes were measured at varying levels of brightness. from this, it was possible to develop an experimental formula expressing the relationship between pupil size and brightness. Next, the results of the first experiment were corrected by function and using analysis of variance (ANOVA) showed that brightness did not have an effect. Therefore, a method for removing the effects of brightness upon pupil size as a function of interest in patterns presented at several levels of brightness. corrected pupil sizes correlated to pupil sizes for patterns presented at the same levels of brightness.
Pupil size is well known to change in response to brightness and mental activities, but the correlation between these two reactions in unclear. First, pupillary changes were measured at various brightness, with and without audio instructions which stimulate mental activities. ANOVA found that there is no interaction between pupillary changes due to brightness effects and mental activities. Second, patterns were presented to subjects at various brightness after initially being presented at a constant brightness. The correction function is obtained that controls pupil size according to the brightness level. Subject interest in each pattern is evaluated from the pupillary changes as corrected by this function.
This paper describes influence of audio circumstances upon reading speed and pupil size. In the experiment, reading speed and pupil size was measured under four audio circumstances: silence, back ground music, TV commercial sounds and white noise. The results showed that reading speed under silence was faster than others and one under TV commercial sound was slower, average pupil size under silence was also larger than others and one under white noise was smaller. The authors report the relationship between reading speed and subjects' behavior which indicate consciousness to sounds. The influence of turning up sound pressure level upon reading speed was measured. It was found influence under back ground music was small, one under white noise was bigger.
Pupillary changes were measured in response to audio instructions while a subject looked at a visual pattern. Results shows that audio instructions cause the pupil to dilate and that pupillary changes are directly related to mental activity. The same results are found when audio instructions are given without presentation of visual patterns. Pupillary changes which occur when a pattern is looked at fixedly are discussed, and subjects can be categorized by their types of pupillary changes. It is suggested that the pupillary response is an indicator of mental activity.