In this paper, we show current status of gathering large scale human activity information using mobile sensor devices. If human activity can be objectively measured, we can expect various applications, such as health care and agriculture. We developed an activity information gathering system using mobile sensor devices with three axis accelerometers, and gathered about 10,000 activity data of 170 people during 5 months so far. As a result, we found specific activities which are easily gathered and ones which are not. Moreover, we found some problems of motivation measures.

SUMMARYWe consider RFID bookshelves, which detect the location of books using RFID. An RFID bookshelf has the antennas of RFID readers in the boards, and detects the location of an RFID tag attached to a book. However, the accuracy is not good with the experience of the existing system, and sometimes reads the tag of the next or even further area. In this paper, we propose a method to improve the location detection using naive Bayes classifer, and show the experimental result. We obtained 78.6% of F-measure for total 12658 instances, and show the advantage against the straightforward approach of calculating the center of gravity of the read readers. More importantly, we show the performance is less dependent of a change of layouts and a difference of books by leave-1-layout/book-out cross validation. This is favorable for the feasibility in library operation.