权益分类	功能权益	普通用户	{{item.name}}会员
{{category.name}}	{{benefitItem.name}}

A study on the cross-media data retrieval based on Kansei

基于感性的跨媒体数据检索研究

基本信息

批准号：
16500067
负责人：
HOCHIN Teruhisa
金额：
$ 2.05万
依托单位：
Osaka Prefecture University
依托单位国家：
日本
项目类别：
Grant-in-Aid for Scientific Research (C)
财政年份：
2004
资助国家：
日本
起止时间：
2004 至 2005
项目状态：
已结题

项目摘要

This aim of this work is the accuracy improvement of multimedia retrieval system based on human impression. The first approach is the automatic extraction of the keywords from Web pages. The keywords include adjectives and adjectival verbs as well as nouns. The experimental result shows that the nouns extracted can be permitted as keywords of the Web pages. On the other hand, the adjectives and adjectival verbs can not be admitted as the keywords. This may be caused by the fact that there are few adjectives and adjectival verbs in a Web page. The second approach of this work includes the personal adaptation and the feeling adaptation of the Kansei retrieval system. To make the personal adaptation easy, a method of decreasing the user's burden in learning his/her taste is developed. This method is based on the grouping of the multimedia data and that of the users. The representative multimedia data and the impression word pairs are decided based on the groups. It is experimentally clarified that the personal adaptation is really attained, and the improvement is required. Effects of feeling on sensitivity to pictures and behavior of Kansei multimedia data retrieval is experimentally evaluated. This evaluation clarified several phenomena. The first is that the feeling affects the sensitivity. Two types of behavior can be observed : the emphasizing behavior and the inhibiting one. The type of behavior affects the sensitivity. The two types of behavior make the retrieval method easy. That is, the retrieval can start only by specifying the user's feeling.

本文的目的是提高基于人的印象的多媒体检索系统的准确性。第一种方法是从Web页面中自动提取关键字。关键词包括形容词、形容词动词和名词。实验结果表明，所提取的名词可以作为网页的关键词。另一方面，形容词和形容词动词不能作为关键词。这可能是由于Web页面中很少有形容词和形容词动词。本研究的第二种方法包括感性检索系统的个人适应和感觉适应。为了使个人适应更容易，开发了一种减少用户学习他/她的口味负担的方法。该方法基于多媒体数据和用户数据的分组。在分组的基础上确定具有代表性的多媒体数据和印象词对。实验结果表明，个体适应能力确实达到了一定程度，需要进一步提高。实验评估了感觉对感性多媒体数据检索的图像敏感性和行为的影响。这一评价澄清了几个现象。第一种是感觉影响敏感性。可以观察到两种类型的行为：强调行为和抑制行为。行为的类型会影响灵敏度。这两种类型的行为使检索方法变得简单。也就是说，检索只能从指定用户的感觉开始。