Recognition of Loneliness of the Elderly People in Ubiquitous Computing Environment - Advances in Computational Science and Engineering - page 62

As the next method, there is a method using RFID tag. Since price of RFID tag has
been lowered a lot due to recent development and industrialization of RFID
technology, this method has advantage of lower price and easy installation in general.
And it is relatively easy to recognize and distinguish Object and also easy to
distinguish multiple users. As disadvantage, due to characteristics of water or RFID,
recognition is not established near metal or water due to frequency interruption and
there can be problems of collision in case density of tag is increased. And since tag is
detected using bracelet type Reader, it is inconvenient to use in places such as bed
room or bath room.
Human Activity Recognition study developed by Intel is a representative study of
this method. In this study, 14 behaviors to which Caregiver pay attention are
recognized and unlike STAR project, probability of behavior is determined by mining
web information in order to avoid labeling through manual works [4]. However,
difference and habit of each user were not considered and it is impossible to predict
next Activity if starting behavior of Activity is changed. In addition, since parameters
of time were fixed, it can be difficult to recognize more realistic behaviors.
The two big behavior recognition studies were reviewed in previous part; all have
used machine learning techniques based on probability. And for learning of machine
learning, supervised learning was used but it can become problems since lots of time
and cost are required for users to label actually large amounts of data one by one and
also behavior pattern of each individual can be different.
In addition, reviewing recognition scope and object, STAR project is simple and
sequential behavior and a combination of those behaviors and in the study of Intel; it
is to recognize behaviors that general people or caretakers pay attention. Like this,
most studies using binary sensor are focused on recognizing behaviors but in this
study, it focuses on recognizing situations in which the aged feel loneliness instead of
sequential behaviors.
Actually, it is very difficult to correctly find emotion of people like loneliness with
binary sensor such as RFID. For this, it is necessary to have modeling regarding
accurate behavior or emotion of people and it is field that more studies are required
yet. Therefore, in this system, method to ratiocinate lonely patterns of the aged using
rule in XML form was used. Advantage of this method, it is possible to add it to rule
in XML form when additional modeling regarding loneliness is realized in the future
after making a big Frame of the whole. Since it is not necessary to make a module to
detect new lonely situation every time and it only needs to add rule. Therefore, it is
not necessary to modify internal code. Also, it is easy to add new rules for difference
and habit of each user and it is also relatively stronger for noises compared to the
existing methods since time parameter is used.
3 Scenario for the Loneliness Inference System
The scenario for the Loneliness inference system for elderly that is suggested in this
paper is as follows.
① Attach RFID tags to furniture or other objects in the user's house.
② The user should wear the RFID reader on their wrist.
③ The user should go about their routine activities.