Free download thesis
I. INTRODUCTION
Falls in elderly is a major public problem. Several studies
have shown [1] that elderly people experience at least one
fall every year. Moreover, falls are the main cause of
accidental death in older adults aged 65 or more [2].
Therefore, fall detection has attacted a large attention of
researchers as well as industial. A huge number of material,
equipment and fall detection methods have been proposed.
Among these methods, vision based method has a great
advantage because it does not require the elderly people to
wear a specifc equipment.
Recently, Kinect device has been released. The use of
this sensor for health care application in general and fall
detection in particular has two main advantages. Firstly, this
sensor provides rich information (color, depth and skeleton)
in comparison with conventional cameras and therefore gives
a better representation of fall. Secondly, this device allows
developing a solution for protecting privacy - major issue i
health care application by using only depth and skeleton
information.
In this paper, we present a novel fall detection system
based on the Kinect device. The originalities of this system
are two-fold. Firstly, based on the observation that using all
joints is not pertinent and robust to represent human posture
because in several human postures the Kinect is not able to
track correctly all joints, we defne and compute tree features (distance, angle, velocity) on only several important
joints. Secondly, in order to distinguish fall with the others
activities such as lying, we propose to use SVM (Support
Vector Machine). The experimental results show that the
proposed system is capable of detecting falls accurately and
robustly.
The remaining of this paper is organized as follows. The
section II gives a brief survey on vision-based fall detection
approaches in general and Kinect-based fall detection i
particular. Then, our proposed method is described in detail
in Section III. Section IV presents experimental results as
well as discussions. Finally, we give some conclusions and
fture works in Section V.
II. RELATED WORKS
The fall of a person can be described as the rapid
change fom upright/sitting position to the reclining or
almost lying position. The fall is not a controlled movement
like lying for example. According to [3], the fall may be
divided into four phases: prefall, critical, postfall and
recovery.
Over the last years, there are a number of works that
have been proposed for fall detection. These works can be
classifed according to whether they focus on direct
detection of the critical phase/impact shock or postfall
phase. In this paper, we do not try to do an exhaustive
survey of all state of the art works for fall detection that is
the scope of the others papers [3]. We present a brief
summary of vision-based fall detection. As our work is
based on Kinect device, we try to analyze the related works
for fall detection using Kinect.
Since Kinect device provides color, depth image and
skeleton information. The Kinect-based fall detection work
can be classifed according information source. The works
belonging to the frst category explores color and depth
information while the second category attempts to use
skeleton information to detect the fall.
Concering fall detection using depth information,
Rougier et al. [4] introduce a Kinect-based fall detection
system. In this system, a fall will be recognized using two
parameters: the distance between subject's centroid and the
foor and the velocity of the center of mass. In this work, the
Do Drive thay đổi chính sách, nên một số link cũ yêu cầu duyệt download. các bạn chỉ cần làm theo hướng dẫn.
Password giải nén nếu cần: ket-noi.com | Bấm trực tiếp vào Link để tải:
I. INTRODUCTION
Falls in elderly is a major public problem. Several studies
have shown [1] that elderly people experience at least one
fall every year. Moreover, falls are the main cause of
accidental death in older adults aged 65 or more [2].
Therefore, fall detection has attacted a large attention of
researchers as well as industial. A huge number of material,
equipment and fall detection methods have been proposed.
Among these methods, vision based method has a great
advantage because it does not require the elderly people to
wear a specifc equipment.
Recently, Kinect device has been released. The use of
this sensor for health care application in general and fall
detection in particular has two main advantages. Firstly, this
sensor provides rich information (color, depth and skeleton)
in comparison with conventional cameras and therefore gives
a better representation of fall. Secondly, this device allows
developing a solution for protecting privacy - major issue i
health care application by using only depth and skeleton
information.
In this paper, we present a novel fall detection system
based on the Kinect device. The originalities of this system
are two-fold. Firstly, based on the observation that using all
joints is not pertinent and robust to represent human posture
because in several human postures the Kinect is not able to
track correctly all joints, we defne and compute tree features (distance, angle, velocity) on only several important
joints. Secondly, in order to distinguish fall with the others
activities such as lying, we propose to use SVM (Support
Vector Machine). The experimental results show that the
proposed system is capable of detecting falls accurately and
robustly.
The remaining of this paper is organized as follows. The
section II gives a brief survey on vision-based fall detection
approaches in general and Kinect-based fall detection i
particular. Then, our proposed method is described in detail
in Section III. Section IV presents experimental results as
well as discussions. Finally, we give some conclusions and
fture works in Section V.
II. RELATED WORKS
The fall of a person can be described as the rapid
change fom upright/sitting position to the reclining or
almost lying position. The fall is not a controlled movement
like lying for example. According to [3], the fall may be
divided into four phases: prefall, critical, postfall and
recovery.
Over the last years, there are a number of works that
have been proposed for fall detection. These works can be
classifed according to whether they focus on direct
detection of the critical phase/impact shock or postfall
phase. In this paper, we do not try to do an exhaustive
survey of all state of the art works for fall detection that is
the scope of the others papers [3]. We present a brief
summary of vision-based fall detection. As our work is
based on Kinect device, we try to analyze the related works
for fall detection using Kinect.
Since Kinect device provides color, depth image and
skeleton information. The Kinect-based fall detection work
can be classifed according information source. The works
belonging to the frst category explores color and depth
information while the second category attempts to use
skeleton information to detect the fall.
Concering fall detection using depth information,
Rougier et al. [4] introduce a Kinect-based fall detection
system. In this system, a fall will be recognized using two
parameters: the distance between subject's centroid and the
foor and the velocity of the center of mass. In this work, the
Do Drive thay đổi chính sách, nên một số link cũ yêu cầu duyệt download. các bạn chỉ cần làm theo hướng dẫn.
Password giải nén nếu cần: ket-noi.com | Bấm trực tiếp vào Link để tải:
You must be registered for see links