Recognition of Loneliness of the Elderly People in Ubiquitous Computing Environment - Advances in Computational Science and Engineering - page 61

and other living supplies within homes and present status information of the aged was
arranged to be collected through RFID reader attached to wrists of users for this
purpose. Using information obtained like this, whether the aged are in lonely situation
presently is ratiocinated and if they are in lonely situation, images helpful to alleviate
loneliness were provided to the aged.
However, it was impossible to quantify since loneliness is feeling of people and
lonely situation is different significantly per each person and also it is needed to
consider habit of each person. To solve these problems, flexibility of system is
required to add or modify situation in which each person feels loneliness. Therefore,
in this system, method that ratiocinates lonely pattern of the aged using rule in XML
form was suggested. Strong advantage of this method is that it is possible to add it to
rule in XML format when additional modeling of lonely situation is realized in the
future after a big Frame is made as the whole. Since only Rule is required to be added,
it is not necessary to create module to detect new lonely situation every time and
therefore, it is not necessary to modify internal code.
In the next section, related studies were reviewed and lonely situation was defined
and was expressed as Rule in section 3. And in section 4, through Jess ratiocination
engine, with the Rule defined in section 3, method of ratiocination was explained and
in section 5, overall system structure was explained [1]. Finally, in section 6, direction
and improvement points in the future were mentioned and conclusion was described.
2 Related Work
Behavior recognition studies can be classified into two kinds except present video
sensor.
First, there is a system built using binary sensor of the existing theft prevention
system or fire alarm system and as advantage of this system, it is cheap in price since
it uses the existing theft prevention system and it is also possible to recognize by
using near water or metal. In addition, it does not give burdens to users since users
only need to wear relatively simple devices. On the contrary, as disadvantage, for
homes without installation of theft prevention system, it is a very expensive method
and it has limitation in recognizing delicate Activity because it is difficult to increase
density of sensor due to the characteristics of binary sensors. In addition, it is
relatively difficult to recognize and distinguish Objects compared to RFID
technology. And as one of the biggest weaknesses is that it is very difficult to
distinguish in case when there are many users.
STAR project, which is a representative study of this method, RFID tag, motion
detect sensor, Break Beam sensor, contact sensor, and pressure sensor are installed in
usual homes and through these, not only behaviors of users are recognized but also
episode can be induced using recognized behaviors [2, 3]. However, tremendous
amount of data association are generated as it has considered even environment with
many users. So in order to solve this problem, accuracy has been increased using
particle filter. However, as the number of users increase, accuracy is reduced
drastically and it can be problems in handling on real time as collection time of
parameter estimation is long.