Summarization and Indexing of Home and Wearable Video

通过电子手段捕捉生活日志，使我们能够详细记录我们的日常生活。到目前为止，我们只依靠人类的记忆来记录和记住我们的日常经历。我们往往很快就会忘记我们经历的细节。我们用视频和照片记录特殊的场合，如旅行和家庭活动。排除这些特殊事件，我们很少记录日常生活中的主要经历，我们最多做的是写一篇简短的日记。随着可穿戴、无处不在、移动多媒体相关技术的发展，我们相信我们将能够自动捕获和记录我们的日常体验。在这个项目中，我们一直在研究通过各种传感器，如GPS(位置)、陀螺仪、加速度传感器(运动)、生理传感器(脑电波)、注释、文档、电子邮件和网页，通过音频和视频捕获生活日志。检索是生命日志系统中最重要的问题。捕获的数据量非常大，问题是要找到或浏览这些数据，以便我们可以查看我们体验的细节。记住我们经历的上下文比记住它们的细节更容易，所以上下文可以为索引信息提供有价值的关键字。我们一直在开发一个基于上下文的检索系统，通过该系统，我们可以根据时间、位置、人的行为等参数，以及通过注释、文档、电子邮件和网页的词语来导航海量的视频数据。我们可以通过包含在城镇数据库中的商店和商店的名称来搜索视频。

Capturing a life log by electronic means enables us to record our daily life in detail. To date we have only relied on our human memory to record and remember our daily experiences. We tend to quickly lose the details of our experiences. We record special occasions such as travels and family events with videos and photos. Excluding such special events, we rarely record the daily experiences that are the major part of our life ; what we do at most is writing a short diary. Because of developing technologies related to wearable, ubiquitous, mobile multimedia, we believe that we will be able to automatically capture and record our daily experiences.In this project, We have been investigating capturing a life log by audio and video with various sensors such as a GPS (location), gyros, acceleration sensors (motion), physiological sensors (brain wave), annotations, documents, emails, and web pages. Retrieval is the most important problem for the life log system. The amount of data captured is very large, and the problem is to find or navigate through the data so that we can watch the details of our experiences. Remembering the context of our experiences is easier than remembering details of them, so the context can provide valuable keys for indexing information. We have been developing a context-based retrieval system, by which we can navigate the huge amount of video data by parameters such as time, location, a person's behavior, and through words of annotations, documents, emails and web pages. We can search the video by the names of shops and stores that are contained in the town database.