抄録
There are many lifelog researches to give services with personal daily digitized information.. To realize lifelog services, the labeling technology to enormous quantity of information is necessary. In this paper, we propose a versatile labeling technology focused on the unordinary/ordinary situation of daily life. Our approach regards the difference of data distribution collected in each action. And based on the time-oriented features of the data, we brought in smoothing process, and label-modify-process. We evaluated the accuracy of our approach with heart rate sensor data. The evaluation was conducted on three subjects. Then, we comfirmed more 13.5 % accuracy improvement with above two processes, than without them.
