02-10-2012, 04:18 PM
Detection of Abnormal Living Patterns for Elderly Living Alone Using Support Vector Data Description
[attachment=35053]
INTRODUCTION
RECENTLY, the average life span and proportion of the
elderly population relative to the total population has increased,
and continues to increase, in developed countries. By
2020, most of the developed countries will consist of an aged
society or super-aged society, and Korea is the fastest aging
country in the world [1]. In addition, the number of elderly people
living alone has increased in Korea. In 1998, 20.1% of the
entire elderly population was living alone, and this percentage
increased up to 24.6% in 2004.
MEASUREMENT SYSTEM
The myriad challenges solitary elderly people face is more
critical in rural communities. The portion of elderly people living
alone is greater in rural areas, where the percent of elderly
people living alone in rural areas is 28.4% more than urban areas
and 18.4% in Korea [18]. In this study, we first established a
ubiquitous healthcare house in 2008 at Jeung–Pyeong city as a
pilot study to test an IR motion sensor system so that we could
measure daily activity information of elderly subjects living
alone. Jeung–Pyeong is a typical rural city that is 105 km away
from Seoul and has a population of 31 483 residents. There are
more than a hundred elderly people living in an isolated area
away from downtown and these people have suffered from the
healthcare problems due to living alone. Thus, it was necessary
to monitor the daily activity of the elderly living in this area to
detect abnormal behavior patterns early on and to prevent them
from worsening.
ABNORMALITY DETECTION
The proposed system was designed to monitor the daily life
of the elderly in his/her home. It was conceived to estimate normal
behavior patterns of daily living and to automatically detect
unusual behavior patterns. Additionally, the elderly health monitoring
system should be designed so as to not disturb their
ordinary activity and to maximize their privacy while accurately
detecting irregular situations and behaviors. Therefore,
we found it necessary to analyze the elder’s behavior pattern
simultaneously. The real-time behavior analysis program was
implemented with Windows programming using Visual Studio
2008 (Microsoft). MATLAB (MathWorks) software was used
for signal processing.
RESULTS
Three Features for Daily Living Pattern Classification
The main idea of this study was to use three features to assess
the daily living pattern of the elderly and develop an abnormality
detection method using SVDD classification. Fig. 10 shows
the scatter plots of three features from one subject over 24 h.
The feature values were calculated every hour, and the calculated
feature points were scattered in 3-D space. Fig. 10(a)–(f)
shows the scattered feature values from 0:00 to 4:00, 4:00 to
8:00, 8:00 to 12:00, 12:00 to 16:00, 16:00 to 20:00, and 20:00
to 24:00, respectively. In the early morning between 0:00 and
4:00 A.M. [see Fig. 10(a) and (b)], most of the feature values
were located along the z-axis, which is the NRI feature axis.
[attachment=35053]
INTRODUCTION
RECENTLY, the average life span and proportion of the
elderly population relative to the total population has increased,
and continues to increase, in developed countries. By
2020, most of the developed countries will consist of an aged
society or super-aged society, and Korea is the fastest aging
country in the world [1]. In addition, the number of elderly people
living alone has increased in Korea. In 1998, 20.1% of the
entire elderly population was living alone, and this percentage
increased up to 24.6% in 2004.
MEASUREMENT SYSTEM
The myriad challenges solitary elderly people face is more
critical in rural communities. The portion of elderly people living
alone is greater in rural areas, where the percent of elderly
people living alone in rural areas is 28.4% more than urban areas
and 18.4% in Korea [18]. In this study, we first established a
ubiquitous healthcare house in 2008 at Jeung–Pyeong city as a
pilot study to test an IR motion sensor system so that we could
measure daily activity information of elderly subjects living
alone. Jeung–Pyeong is a typical rural city that is 105 km away
from Seoul and has a population of 31 483 residents. There are
more than a hundred elderly people living in an isolated area
away from downtown and these people have suffered from the
healthcare problems due to living alone. Thus, it was necessary
to monitor the daily activity of the elderly living in this area to
detect abnormal behavior patterns early on and to prevent them
from worsening.
ABNORMALITY DETECTION
The proposed system was designed to monitor the daily life
of the elderly in his/her home. It was conceived to estimate normal
behavior patterns of daily living and to automatically detect
unusual behavior patterns. Additionally, the elderly health monitoring
system should be designed so as to not disturb their
ordinary activity and to maximize their privacy while accurately
detecting irregular situations and behaviors. Therefore,
we found it necessary to analyze the elder’s behavior pattern
simultaneously. The real-time behavior analysis program was
implemented with Windows programming using Visual Studio
2008 (Microsoft). MATLAB (MathWorks) software was used
for signal processing.
RESULTS
Three Features for Daily Living Pattern Classification
The main idea of this study was to use three features to assess
the daily living pattern of the elderly and develop an abnormality
detection method using SVDD classification. Fig. 10 shows
the scatter plots of three features from one subject over 24 h.
The feature values were calculated every hour, and the calculated
feature points were scattered in 3-D space. Fig. 10(a)–(f)
shows the scattered feature values from 0:00 to 4:00, 4:00 to
8:00, 8:00 to 12:00, 12:00 to 16:00, 16:00 to 20:00, and 20:00
to 24:00, respectively. In the early morning between 0:00 and
4:00 A.M. [see Fig. 10(a) and (b)], most of the feature values
were located along the z-axis, which is the NRI feature axis.