紀要論文

ライフログ分析による行動特徴抽出及びイベント検出

The Lifelog Analyses for Activity Pattern Extraction and Event Detection

jpn

18810667

https://doi.org/10.15002/00010532

法政大学大学院情報科学研究科

9

119

124

2014-03

Version of Record

Abstract— Recently,obtaining lifelog becomes much easier and they can be acquired at any time, thanks to the explosive spread of smartphones and the downsized wearable logging devices. From lifelog it is possible to examine user's activities and life styles. In a smartphone there are various kinds of embedded sensors, e.g., GPS, acceleration sensor, luminance sensors and so on, that can be used to capture lifelog. These sensors make smartphones as a valuable life logging tool. In addition, smart phones can be connected to the Internet and this makes it possible to obtain logs when using Web applications, for instance. Web browsing histories and SNS logs like Twitter. There are a lot of researches for analyzing smartphone logs, for examples, the extraction of user's activity using location, date and people around the user, and the analysis of user's activity history by detecting Bluetooth devices in user's circumstance. This research is aiming to mine various kinds of smartphone logs including locations from GPS, pictures from a camera, application using histories, text messages from SNS (Twitter), and number of steps as exercise history from the wearable Jawbone device. Most people tend to spend time under certain patterns or rules based on day and time. But on holidays, people may visit somewhere new, or go out and hang out with some friends or more, or have a trip to far places. In our research an important step is to find features by analyzing the log's occurrences based on day and date. From the result of occurrence and feature analyses of multiple logs, we find out the user's daily life patterns, by which, we can further detect special events happened in these days with different life patters. These events would be valuable to support user's life recall.

Article

AA12222297