Professional Documents
Culture Documents
net/publication/303823363
CITATION READS
1 27
6 authors, including:
Liming Chen
De Montfort University
176 PUBLICATIONS 2,482 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
A Conceptual framework for Adaptive User Interfaces for older adults View project
All content following this page was uploaded by Liming Chen on 01 January 2018.
1 Introduction
Fall accident is not only a severe threat to public health, but even fatal for poor
physical elderly people with walking difficulties. With the social aging problems
getting worse these days [7], the number of elderly people living alone gradually
increased and most of them could not get good health services [10, 18]. The elderly
people are usually living alone, which gives rise to risk of fall accident. According
to statistics, more than 33% of people aged 65 years or older have one fall per year
[15]. Sometimes, it is not the fall but the subsequent injury is the threat to the elderly
people [11], and almost 62% of injury-related hospitalizations for seniors are the
results from falling [3]. Sometimes the subsequent injuries like cerebral hemorrhage
happen and they will be lethal if timely rescues are unavailable [8]. Therefore, an
1
2 Jianfei Sun, Zumin Wang, Liming Chen, Baofeng Wang, Changqing Ji and Shuai Tao
accurate and real-time fall detection method will make timely rescue possible that
avoids the subsequent injury and danger.
At present, there are several types of method based on different principles can
realize fall monitoring, which were shown but not limited to the contents in Figure
1.
Fig. 1 Three types of method based on different principles for fall detection.
First of all, the vision-based monitoring type, for example uses a MapCam (omni-
camera) to capture images and performs image processing over the images [14], of
which fall monitoring success rate closed to 80%. Wong obtained the rate of 86.19%
in detecting the human faint event using thermal cameras in indoor environment
[21]. An unsupervised method was proposed in [16] for detecting abnormal activ-
ity using the fusion of some simple features. The detection rate of falls was about
66.67%. Williams used a distributed network of smart cameras to detect and localize
falls with a high accurate of 97.5% [20]. But these methods can be used only for in-
door fall monitoring. Furthermore, due to the application of HD panoramic camera
as the sensor [17], it is likely to involve privacy problems. Even without the privacy
problem, the camera will bring more or less discomfort to monitored people.
In our previous study, we used an infrared sensor network locating on the ceiling
to monitor the action of people, with the higher monitoring accuracy (up to a max-
imum of 95.14%) compared with the HD panoramic camera based methods, which
could also solve the personal privacy problems (without camera equipments). But
there are still some limitations, such as the indoor application only and its system is
relatively complex, the setup and maintenance is more difficult.
The third type of fall monitoring method is based on the use of wearable devices,
such as the fall sensor fixed on the waist in early time with small volume and high
sensitivity [6, 5]. But the wearing discomfort and other shortcomings severely limit
its development; the fall detection function was embedded into the smart bracelet
and even vest [1, 2], which could break through the indoor limitation. But the limits
of this type of method are that the elderly people usually forget to wear the equip-
ments and it is difficult to maintain the monitoring precise in daily life.
A Plantar Inclinometer Based Approach to Fall Detection in Open Environments 3
Take the most commonly used smart wristband as an example, it is very easy
and convenient for the stylish and curious young people to accept new things and
wear such smart wristband, but for the majority of elder people, watch-like portable
devices can bring inconvenience and some of the elderly due to loss of memory,
inevitably forget to wear some unnecessary devices. To solve this kind of problems,
some research groups tried to place pressure sensor and three-axis acceleration sen-
sor on the head and the waist to monitor fall and they obtained a sensitivity of
95.71% and a specificity of 97.78% [4], but wearing this equipment is very cumber-
some for elder people and everyday wearing will be a burden.
In this research, we applied the plantar axial inclinometer based human fall mon-
itoring method, which could effectively reduce limitations of present methods, pro-
vide an efficient and stable system with no environmental dependency and a wide
application without violation of privacy issues.
In summary, our fall monitoring system and method the advantages of simple
structure and low computational complexity, so we can easily integrate it into shoes
or insoles. The elderly people will remember to wear them as long as they simply
take on shoes. So compared with other fall detection systems, ours will not only
preserve user privacy but also with low influence and easy applicable for elderly
people.
In the plantar inclinometer, we use the single axis sensor SCA60C − N1000060,
which is a high-precision, reliable and low-price analog sensor. The photograph of
the sensor module is shown in figure 2. The sensor of SCA60C − N1000060 is ap-
plied widely in the research of intelligent control and abnormal alarm [13, 19]. Due
to the character of single axis, in this study, for correctly reflecting the plantar an-
gle, we need to use two SCA60C − N1000060 in the sensor module (shown in figure
2). In the sensor module, two sensors are orthogonal to each other for detecting the
plantar angle data in two axes.
The sensor module can detect the plantar angles from −90◦ to 90◦ , the corre-
sponding output digital quantities are from -0.5 V to 4.5 V. The relationship between
digital value and analog output is shown in figure 3.
In figure 3, we see that the relationship between digital value (voltage) and analog
output (angle) is nonlinear, and similar to sinusoidal vibration, which is shown as
−O f f set
α = arcsin( Vout
Sensitivity ), (1)
where α is the axis angle of the sensor module, Vout is the digital output. When the
angle of the module is 0o (horizontal direction), the value of O f f set is 2.5 V. The
Sensitivity under room temperature is 2 V/g. Therefore, formula (1) can be rewrite
as
4 Jianfei Sun, Zumin Wang, Liming Chen, Baofeng Wang, Changqing Ji and Shuai Tao
Fig. 3 The relationship between digital value and analog output of the sensor module.
Fig. 4 The sensor module mounted onto a pad, the pad sticked under the feet, a volunteer wearing
the pad.
We set the x-axis and y-axis of left and right foot shown in figure 5. Therefore,
when a person stands upright, the angles of x-axis and y-axis will both be 90◦ .
The subject in this test is a student in our laboratory, who is heathy and can walk
normally. We detected the right-foot angles during walking and found that the angles
are almost 90◦ when the person is standing, and fluctuate rapidly during walking.
The y-axis has a larger fluctuate range. The variation trend of angles during walking
in ten seconds is shown in figure 6.
We also investigated the variation of right-foot angles (left-foot is symmetrical)
when the subject falls forward and backward. From figure 7, we see that when falls
happen, the angles of both x-axis and y-axis have strenuous vibrations, which are
very different from the situation of walking. Then plantar angle data were collected
6 Jianfei Sun, Zumin Wang, Liming Chen, Baofeng Wang, Changqing Ji and Shuai Tao
Fig. 6 The variation trend of right-foot angles during walking in ten seconds.
after falls in four different situations: forward, backward, left and right for deter-
mining the thresholds.
Fig. 7 The variation of right-foot angles (left-foot is symmetrical) when the subject falls forward
and backward.
The volunteer falls forward to an air bed, stays still naturally, the plantar angle data
are then collected. The distribution of plantar angles after forward falls 120 times is
shown in figure 8(a) and 8(b).
From figure 8, we see that the variation range of plantar angle is limited after
falls happening. For this reason, we could select some thresholds to recognize falls.
When a person falls forward, there will be a wide variation range of x-axis, but a
narrow variation range of y-axis.
A Plantar Inclinometer Based Approach to Fall Detection in Open Environments 7
Fig. 8 The distribution of plantar angles after forward falls 120 times.
The volunteer falls backward to an air bed, stays still naturally, the plantar angle
data are then collected. The distribution of plantar angles after backward falls 120
times is shown in figure 9(a) and 9(b).
Fig. 9 The distribution of plantar angles after backward falls 120 times.
From figure 9, we see that when a person falls backward, there will be a wide
variation range of x-axis, but a narrow variation range of y-axis. The y-axis angle
direction of backward falls (0◦ -90◦ ) is contrary to forward falls (90◦ -180◦ ).
8 Jianfei Sun, Zumin Wang, Liming Chen, Baofeng Wang, Changqing Ji and Shuai Tao
The volunteer falls to left side to an air bed, stays still naturally, the plantar angle
data are then collected. The distribution of plantar angles after left falls 120 times is
shown in figure 10(a) and 10(b).
Fig. 10 The distribution of plantar angles after left falls 120 times.
From figure 10, we see that the variation range of plantar angle is limited after
falls happening. When a person falls to left side, most of the variation range of x-axis
is 0◦ -120◦ , most of the variation range of y-axis is 0◦ -120◦ .
The volunteer falls to right side to an air bed, stays still naturally, the plantar angle
data are then collected. The distribution of plantar angles after right falls 120 times
is shown in figure 11(a) and 11(b).
From figure 11, we see that the change range of plantar angle is limited after falls
happening. When a person falls to right side, most of the variation range of x-axis
is 60◦ -180◦ , most of the variation range of y-axis is 0◦ -120◦ .
The limitation to the range of plantar angles of falls caused by physiological
structure [9, 12] of people is an important basis for fall detection in this research.
By observing the collected data of plantar angles, we see that, for x-axis or y-axis
under one of the feet, at least one of them is very large or small. Therefore, based
on this rule, we selected some thresholds for fall detection. The thresholds should
satisfy most of the angle values for fall detection. By statistical analysis, we found
A Plantar Inclinometer Based Approach to Fall Detection in Open Environments 9
Fig. 11 The distribution of plantar angles after right falls 120 times.
that all the angles of at least one axis is in the interval of [57◦ ,131◦ ]. Therefore, the
thresholds of plantar angle for fall detection is set to [57◦ ,131◦ ]. We detect falls by
the formula below.
State = [(57◦ ≥ Rx ∨ Rx ≥ 131◦ ) ∨ (57◦ ≥ Ry ∨ Ry ≥ 131◦ )]
∧[(57◦ ≥ Lx ∨ Lx ≥ 131◦ ) ∨ (57◦ ≥ Ly ∨ Ly ≥ 131◦ )] (3)
x
1, ∑ Staten = x
Fall = n=1 (4)
x
0, ∑ Staten < x
n=1
In the experiment, there are five subjects. The information of subjects are shown
in Table 1. Each subject did five groups of experiments for fall detection in the
horizontal plane (forward, backward, left side, right side and free fall). There are
16 falls in each group and totally 80 falls. In addition, We did another five groups
of experiments for fall detection on a slope. There are 4 falls in each group and
totally 20 falls. The number of falls of each subject in the experiment is 100. The
10 Jianfei Sun, Zumin Wang, Liming Chen, Baofeng Wang, Changqing Ji and Shuai Tao
experiment result of each subject is shown in figure 12. The average detection results
of 500 falls of five subjects are shown in figure 13.
From figure 13, we see that there are 427 falls detected successfully. The detec-
tion rate was 85.4%. However, the detection rate of backward falls is only 77%. The
most possible reason is that slope brings bad effects to the detection. It could cause
a low detection rate in the situation of backward fall due to the effect on plantar
angles.
4 Discussion
In this paper, we report a method of fall detection using plantar inclinometer sensor,
which provides us the information of angle variations during walking, and of angle
A Plantar Inclinometer Based Approach to Fall Detection in Open Environments 11
status after a fall. We analyzed the normal range of angle variations during walking,
and selected the thresholds by testing the distribution of plantar angles. We detectd
falls in different situations by five subjects for 500 times and obtained the average
detection rate of 85.4%.
By comparison with some similar researches in Table 2, it is not difficult to find:
the best success rates were yielded by distributed multi-cameras [20], thermal in-
frared matrix solutions [17] and pressure/acceleration sensor solutions [4], the suc-
cess rates were above 95%; while the thermal infrared matrix fall detection belongs
to environment detection scheme, which is similar to the camera fall detection [14]
that is with larger hardware scale, more complicated algorithm and it is difficult to
be integrated on small scale leading to the restriction to interior application.
Fall detection scheme Detection rate The Complexity The difficulty Environment
of hardware of algorithm
Plantar inclinometer 85.4% [this study] Low Low Indoor/Outdoor
Camera 79.8% [14], 66.67% [16] Medium High Indoor
Distributed multi-cameras 97.5% [20] High High Indoor
Thermal camera 86.19% [21] Medium High Indoor
Pressure and acceleration 95.71% [4] Medium High Indoor/Outdoor
Infrared sensor array 95.14% [17] High High Indoor
The integrated waist/hand acceleration sensors are currently popular fall detec-
tion solutions on the market [6, 5, 1, 2, 4], which are characterized by less hardware
complexity, easy to be integrated on small scale and no interior limitation.
The success rate of most of the products is not high except several solutions ap-
plying multiple sensor types. In order to solve this problem, developers always use
12 Jianfei Sun, Zumin Wang, Liming Chen, Baofeng Wang, Changqing Ji and Shuai Tao
more efficient and complex algorithms, however, which will increase the cost of
hardware due to higher sensitivity, and the complexity of software system. Com-
pared with the solutions above, the main advantages of our system are as follows:
high accuracy, simple hardware, easily to be integrated and without the external
equipment and environment surport. However, there are some deficiencies in the
system: lower sensor accuracy and anti-interference ability; the impedance in sensor
is insensitive to vibrations and it is not possible to achieve fast dynamic detection;
extensible function is limited; the cost of the hardware is relatively high.
Although we obtained a high detection rate of falls, our threshold-based method
is relatively naive, and the subjects are all young people in the experiment. To
solve the problems mentioned above, we will apply machine learning techniques
to build classified models of different activities of elders, and to learn users’ pat-
tern/threshold values. In addition, the difference of angle variation between young
people (the subjects in this study) and elders will be investigated. We expect to in-
crease the proformance by using more advanced methods with the same device of
plantar inclinometer sensors in future study.
Acknowledgements This work was supported by the National Natural Science Foundation of
China (Grant No. 61501076)
References
1. Angelini, L., Caon, M., Carrino, S., Bergeron, L., Nyffeler, N., Jean-Mairet, M., Mugellini,
E.: Designing a desirable smart bracelet for older adults. In: Proceedings of the 2013 ACM
conference on Pervasive and ubiquitous computing adjunct publication, pp. 425–434. ACM
(2013)
2. Bourke, A.K., Van de Ven, P.W., Chaya, A.E., OLaighin, G.M., Nelson, J.: Testing of a long-
term fall detection system incorporated into a custom vest for the elderly. In: Engineering in
Medicine and Biology Society, 2008. EMBS 2008. 30th Annual International Conference of
the IEEE, pp. 2844–2847. IEEE (2008)
3. of Canada, P.H.A.: Report on Seniors’ Falls in Canada [electronic Resource]. Division of
Aging and Seniors, Public Health Agency of Canada (2005)
4. Chen, D., Zhang, Y., Feng, W., Li, X.: A wireless real-time fall detecting system based on
barometer and accelerometer. In: Industrial Electronics and Applications (ICIEA), 2012 7th
IEEE Conference on, pp. 1816–1821. IEEE (2012)
5. Chen, G.C., Huang, C.N., Chiang, C.Y., Hsieh, C.J., Chan, C.T.: A reliable fall detection
system based on wearable sensor and signal magnitude area for elderly residents. In: Aging
Friendly Technology for Health and Independence, pp. 267–270. Springer (2010)
6. Chen, J., Kwong, K., Chang, D., Luk, J., Bajcsy, R.: Wearable sensors for reliable fall detec-
tion. In: Engineering in Medicine and Biology Society, 2005. IEEE-EMBS 2005. 27th Annual
International Conference of the, pp. 3551–3554. IEEE (2006)
7. Cockerham, W.C.: This aging society. (1991)
8. Dunning, K., Moomaw, C., Flaherty, M.L., Osborne, J., James, M.L., Worrall, B.B., Woo, D.:
Abstract t p279: Falls after intracerebral hemorrhage. Stroke 46(Suppl 1), ATP279–ATP279
(2015)
9. Elveru RA Rothstein JM, L.R.: Goniometric reliability in a clinical setting. subtalar and ankle
joint measurements. Physical therapy 68(5), 672–677 (1988)
A Plantar Inclinometer Based Approach to Fall Detection in Open Environments 13
10. Gao, L., Cao, X., Zhang, M.: The study on community health education of empty nest elderly.
Engineering 5(10), 137 (2013)
11. Guimaraes, R., Isaacs, B.: Characteristics of the gait in old people who fall. Disability &
Rehabilitation 2(4), 177–180 (1980)
12. Leardini A O’Connor JJ, C.F.G.S.: A geometric model of the human ankle joint. Journal of
Biomechanics 32(6), 585591 (1999)
13. LIU, X.q., KUANG, L.j., LI, J.c., HUANG, J.z.: The design and production of panels control
system based on the angle sensor sca60c. Mechanical & Electrical Engineering Technology
8, 011 (2012)
14. Miaou, S.G., Sung, P.H., Huang, C.Y.: A customized human fall detection system using omni-
camera images and personal information. In: Distributed Diagnosis and Home Healthcare,
2006. D2H2. 1st Transdisciplinary Conference on, pp. 39–42. IEEE (2006)
15. Noury, N.: A smart sensor for the remote follow up of activity and fall detection of the elderly.
In: Microtechnologies in Medicine & Biology 2nd Annual International IEEE-EMB Spe-
cial Topic Conference on, pp. 314–317. IEEE (2002)
16. Rougier, C., Meunier, J., St-Arnaud, A., Rousseau, J.: Monocular 3d head tracking to detect
falls of elderly people. In: Engineering in Medicine and Biology Society, 2006. EMBS’06.
28th Annual International Conference of the IEEE, pp. 6384–6387. IEEE (2006)
17. Tao, S., Kudo, M., Nonaka, H.: Privacy-preserved behavior analysis and fall detection by an
infrared ceiling sensor network. Sensors 12(12), 16,920–16,936 (2012)
18. Tian, Q., Meng, C.: An empirical investigation of rural empty-nesters in chongqing and a
construction of service system. Asian Agricultural Research 5(02) (2013)
19. Wenlong, Z., Qing, G., Baoshan, L.: Design of landslide warning system. In: Measuring
Technology and Mechatronics Automation (ICMTMA), 2011 Third International Conference
on, vol. 1, pp. 974–977. IEEE (2011)
20. Williams, A., Ganesan, D., Hanson, A.: Aging in place: fall detection and localization in a
distributed smart camera network. In: Proceedings of the 15th international conference on
Multimedia, pp. 892–901. ACM (2007)
21. Wong, W.K., Lim, H.L., Loo, C.K., Lim, W.S.: Home alone faint detection surveillance system
using thermal camera. In: Computer Research and Development, 2010 Second International
Conference on, pp. 747–751. IEEE (2010)