Conference PaperPDF Available

A Comparison of Machine Learning Algorithms for Fall Detection using Wearable Sensors

Authors:

Abstract and Figures

The proportion of people 60 years old and above is expected to double globally to reach 22% by 2050. This creates societal challenges such as the increase of age-related illnesses and the need for caregivers. Falls are a major threat for the elderly, often causing serious injuries especially when the fallen person stays on the ground for a long time without assistance. This paper presents the development of a Fall Detection System (FDS) using an accelerometer combined with a gyroscope worn at the waist. Data come from SisFall, a publicly available dataset containing records of Activities of Daily Living and falls. We compared five Machine Learning algorithms. We first applied preprocessing and a feature extraction stage before using five Machine Learning algorithms, allowing us to compare them. Ensemble learning algorithms such as Random Forest and Gradient Boosting have the best performance, with a Sensitivity and Specificity both close to 99%.
Content may be subject to copyright.
A Comparison of Machine Learning Algorithms for
Fall Detection using Wearable Sensors
Nicolas Zurbuchen, Pascal Bruegger
Institute of Complex Systems (iCoSys)
School of Engineering and Architecture of Fribourg Switzerland
HES-SO University of Applied Sciences and Arts Western Switzerland
Fribourg, Switzerland
nicolas.zurbuchen@hes-so.ch, pascal.bruegger@hes-so.ch
Adriana Wilde
Centre for Health Technologies (CHT)
School of Electronics and Computer Science
University of Southampton
Southampton, United Kingdom
agw106@ecs.soton.ac.uk
Abstract—The proportion of people 60 years old and above is
expected to double globally to reach 22% by 2050. This creates
societal challenges such as the increase of age-related illnesses
and the need for caregivers. Falls are a major threat for the
elderly, often causing serious injuries especially when the fallen
person stays on the ground for a long time without assistance.
This paper presents the development of a Fall Detection
System (FDS) using an accelerometer combined with a gyroscope
worn at the waist. Data come from SisFall, a publicly available
dataset containing records of Activities of Daily Living and
falls. We compared five Machine Learning algorithms. We first
applied preprocessing and a feature extraction stage before
using five Machine Learning algorithms, allowing us to compare
them. Ensemble learning algorithms such as Random Forest and
Gradient Boosting have the best performance, with a Sensitivity
and Specificity both close to 99%.
Index Terms—fall detection; wearable sensors; sampling rate;
data preprocessing, feature extraction, machine learning
I. INTRODUCTION
Falls are one of the leading causes of death among the
elderly [1]. Every year, 28% to 35% of the elderly fall at
least once and this rate increases with age [2]. Falls can have
severe physical, psychological and even social consequences.
They can also heavily affect the independent quality of living.
They can result in bruises and swellings, as well as fractures
and traumas [3]. A significant risk is the long-lie. This happens
when an elderly person remains on the ground for a long
duration without being able to call for help. It is associated
with death within the next few months following the accident
[4]. It also affects the elderly’s self-confidence who may
develop the fear of falling syndrome. It leads to anxiety when
performing Activities of Daily Living (ADLs) and can lead to
subsequent falls [1].
Therefore, the elderly must continuously be monitored to
ensure their safety. Families organize visits but these can be
inconvenient and even insufficient. Hiring caregivers or mov-
ing into nursing homes are sometimes not affordable options.
Recent progresses in technology have enabled the development
of Assisted-Living Systems (ALSs) [5]. They can assist the
elderly and provide a safer environment through constant
monitoring while relieving caregivers’ workload. However,
ALSs create other challenges such as privacy concerns and
acceptability issues that need to be addressed [6].
Fall Detection Systems (FDSs) are part of ALSs. Their
goals are to identify falls and notify caregivers so that they
can intervene as fast as possible. However, fall recognition
is challenging from a computational perspective. Falls can be
defined as “the rapid changes from the upright/sitting position
to the reclining or almost lengthened position, but it is not
a controlled movement” [7]. There is a higher acceleration
during falls. Another challenge is that falls can happen in
innumerable scenarios. They may occur anywhere at any time
[3]. Their starting and ending body posture as well as their
direction (e.g. forward, backward) may vary [1]. Hence, FDSs
must cover the whole living area. Their reliability must be
high while minimizing false alarms, all the while respecting
the elderly’s privacy.
In this paper, we developed a reliable FDS by the mean of
wearable sensors (accelerometer and gyroscope) and various
Machine Learning (ML) algorithms. The goal is to compare
lazy,eager and ensemble learning algorithms and assess their
results. We implemented five algorithms and tested them in
the same setup.
The rest of this paper is organized as follows. In section
II, we discuss existing FDSs and highlight their distinctive
features. Section III covers the employed methodology. Sec-
tion IV presents and discusses the obtained results. Finally, we
conclude with a comment on future work in section V.
II. RE LATE D WO RK
Scientists have employed various approaches to implement
FDSs over the past years. They have been classified as
presented in Fig. 1. Each of them has its strengths and
weaknesses. We focus on wearable technologies since we use
this approach. Nevertheless, several survey studies [8], [9]
reported the other methods in more depth.
A. Choice of sensors and sampling rate
Several types of sensors including accelerometers, gyro-
scopes, magnetometers, and tilt sensors have been used to
detect falls. Based on the fall characteristics, most studies, such
as [10]–[13], employed only acceleration measurements. Only
Bourke and Lyons [14] used a single biaxial gyroscope and
measured changes in angular velocity, angular acceleration,
and body angle. Wang et al. [15] employed a heart rate monitor
and discovered that the heart rate increases by 22% after
a fall in people over 40 years old. This demonstrates that
physiological data can be used in such a system. Across these
papers, the sensors’ sampling rate varied within a range from
20 to 1000 Hz. This variation is not small, one having 50
times more samples than the other, seemingly arbitrary.
B. Sensing position
The sensor placement highly affects the detection perfor-
mance. Previous studies [12], [16] demonstrated that better
results are achieved when sensors are placed along the longitu-
dinal axis of the body (e.g. head, chest, waist) when compared
to other placements (e.g. thigh, wrist). The movement of this
axis during a fall is more consistent and steady. However,
this requires to wear a dedicated device on uncommon body
parts which consequently creates inconveniences. For this
reason, other studies [10], [17], [18] used commodities (e.g.
smartphones carried by the thigh, smartwatches worn on the
wrist). These usually do not disturb the users since they already
wear them. However, people tend to take these devices off
when they are at home which makes the FDS useless. Another
method is to combine various sensing positions. ¨
Ozdemir et al.
[19] developed a system consisting of six wearable devices that
are all used together. The problem is that the elderly already
have acceptability issues with one device, let alone six.
C. Algorithms
There are two categories of algorithms: threshold-based and
ML-based. Threshold algorithms simply define limit values,
outside of which, a fall is detected.. They have often been
sufficient but they tend to produce false alarms especially with
fall-like activities such as sitting abruptly [14]. To compensate,
these studies [11], [20] added simple posture and pattern
Fig. 1. Classification of Fall Detection System approaches.
Fig. 2. General architecture of Fall Detection Systems.
recognition algorithms that detect changes in body posture and
level of activity. This improves the detection’s robustness while
keeping a low computational complexity. However, it may still
fail during specific falls and ADLs. For example, Sucerquia et
al. [21] used a threshold-based classification over their dataset
SisFall, achieving 96% accuracy.
ML algorithms automatically learn patterns based on data,
and very commonly include feature extraction. They require
more computational power and are complex to optimize but
produce improved results. Most of the studies such as [10],
[11] employed a supervised learning technique. Common algo-
rithms are k-Nearest Neighbor [19], Support Vector Machine
[19] and Artificial Neural Network [10], [19]. Yuwono et al.
[13] used unsupervised learning which works with clusters.
This is a compelling solution because it does not require
labelled data. Deep Learning algorithms are nowadays very
popular and achieve promising results in various fields. Musci
et al. [23] employed Recurrent Neural Networks to detect falls.
They used a publicly available dataset [21] and outperformed
the paper’s results.
D. Strengths and weaknesses
Wearable technologies have several advantages. They are
relatively inexpensive, can operate anywhere and require little
computational power all of it with minimal intrusion compared
to other approaches, such as environmental monitoring [16].
They can also identify the wearer and get precise measure-
ments. However, they may create discomfort due to their
size and intrusiveness. The main disadvantage is their human
dependency. These sensors must have enough battery and be
worn to work properly. Furthermore, the elderly may have a
cognitive impairment and thus, may forget to wear the sensor.
III. METHODOLOGY
Our FDS is based on a common pipeline (Fig. 2) which has
been seen in the literature [19]. This pipeline is a common
practice when working with ML algorithms. We first acquire
raw data using various sensors and convert them into discrete
values. We then preprocess the raw data to remove measuring
errors which can badly affect the performance. Afterwards,
we construct and extract meaningful information in a vector.
Finally, we train and evaluate our ML algorithm to distinguish
falls from ADLs.
TABLE I
DETAI LS O F THE ACTIVITIES OF DA ILY LIV ING A ND FA LLS C ON TAINE D
IN T HE SisFall DATASET [21].
Activity Duration
[s]
Walking slowly 100
Walking quickly 100
Jogging slowly 100
Jogging quickly 100
Walking upstairs and downstairs slowly 25
Walking upstairs and downstairs quickly 25
Slowly sit and get up in a half-height chair 12
Quickly sit and get up in a half-height chair 12
Slowly sit and get up in a low-height chair 12
Quickly sit and get up in a low-height chair 12
Sitting, trying to get up, and collapse into a chair 12
Sitting, lying slowly, wait a moment, and sit again 12
Sitting, lying quickly, wait a moment, and sit again 12
Changing position while lying (back-lateral-back) 12
Standing, slowly bending at knees, and getting up 12
Standing, slowly bending w/o knees, and getting up 12
Standing, get into and get out of a car 25
Stumble while walking 12
Gently jump without falling (to reach a high object) 12
Fall forward while walking, caused by a slip 15
Fall backward while walking, caused by a slip 15
Lateral fall while walking, caused by a slip 15
Fall forward while walking, caused by a trip 15
Fall forward while jogging, caused by a trip 15
Vertical fall while walking, caused by fainting 15
Fall while walking with damping, caused by fainting 15
Fall forward when trying to get up 15
Lateral fall when trying to get up 15
Fall forward when trying to sit down 15
Fall backward when trying to sit down 15
Lateral fall when trying to sit down 15
Fall forward while sitting, caused by fainting 15
Fall backward while sitting, caused by fainting 15
Lateral fall while sitting, caused by fainting 15
A. Dataset
We used a publicly available dataset named SisFall [21]. We
selected this dataset over others because of its high quality.
We assessed this quality with various criteria, namely the
size of the dataset and the diversity of subjects in terms of
age, gender, weight, and height. We also took into account
the number of falls and ADLs performed by each subject. In
the SisFall dataset, two triaxial accelerometers (ADXL345 and
MMA8451Q) and a triaxial gyroscope (ITG3200) were used
at a sampling rate of 200 Hz. We decided not to use the data of
the second accelerometer (MMA8451Q) because usual setups
only have a single accelerometer.
Twenty-three young people (19 to 30 years old) performed
15 types of falls and 19 types of ADLs including fall-like
activities. Fifteen elderly people (60 to 75 years old) also
performed the same ADLs for more authenticity. There were
five trials per activity except for the walking and jogging
activities, each of which had only one trial (See Table I).
Hence, SisFall contains a total of 4505 records including 2707
ADLs and 1798 falls, making it unbalanced. A total of 38
people including 19 women and 19 men participated. Table I
lists the falls and ADLs and their duration.
B. Data preprocessing
The SisFall dataset required minimal preprocessing. We
started by equalizing the duration of each record, by equally
cutting (top and tail in equal measure) reducing the length to
10 seconds. We chose 10 seconds to remove outliers induced
by the fall experiment, whilst preserving the fall within each
record.
Regarding the two walking and two jogging activities, which
only have one trial (Table I), we extracted 5 times 10 seconds
for each record. We did this to have the same number of trials
per activity. We selected five windows with no overlap along
each record as follows:
1) From 5 to 15 seconds
2) From 25 to 35 seconds
3) From 45 to 55 seconds
4) From 65 to 75 seconds
5) From 85 to 95 seconds
C. Feature extraction
We then extracted meaningful information from the pre-
processed data. This process helps extracting information that
better characterize each activity. A common practice, when
working with time series, is to extract time and frequency
domains features [19]. A time-domain feature extracted widely
in the literature [10], [18], [19] is the norm of a sample (1).
Norm =pX2+Y2+Z2(1)
We calculated the norm of acceleration and rotation mea-
sures. However, this feature alone is not sufficient to allow a
robust fall detection. In the case of fall-like activities (e.g.
faster movement), this feature would probably be mislead-
ing. Thus, we also extracted time-domain features such as
the variance,standard deviation,mean,median,maximum,
minimum,delta,25th centile, and 75th centile. Additionally,
we extracted frequency-domain features, using a Fast Fourier
Transform (FFT) and we extracted two features: the power
spectral density and the power spectral entropy.
Table II summarizes the selected features. We extracted
them for each axis of each sensor (3 axes, 2 sensors) but also
for each sensor norm. This results in a feature vector of 88
features per record.
Finally, we normalized the extracted features to rescale the
data to a common scale. This gives more influence to data
with small values which can be neglected depending on the
employed algorithm. In this work, we used the common min-
max normalization (2) which scales the values between 0 and
1 included.
min max[0,1] x0=xmin(x)
max(x)min(x)(2)
D. Classification algorithm
We selected five different types of ML algorithms among
the most widely used ones. The following list shortly describes
them:
TABLE II
LIST OF EXTRACTED TIME AND FREQUENCY DOMAINS FEATURES.
Feature Domain
Variance Time
Standard deviation (STD) Time
Mean Time
Median Time
Maximum Time
Minimum Time
Delta (peak-to-peak) Time
25th Centile Time
75th Centile Time
Power Spectral Density (PSD) Frequency
Power Spectral Entropy (PSE) Frequency
k-Nearest Neighbour (KNN) is a simple and popular yet
effective algorithm. Data are classified by a majority
vote with the class most represented among its k-closest
neighbours. It is a lazy learning type of algorithm because
almost no work is done until a prediction. KNN has
previously been used for fall detection and produced
promising results [19].
Support Vector Machine (SVM) is also commonly used
in various tasks. It tries to find the best hyperplane
which maximises the margins between each class. It
is an eager learning algorithm because it works a lot
during the training stage, building a model, in this case,
a hyperplane. It has often been used to detect falls either
with wearable, [19] where it has also produced promising
results, or video-based systems.
Decision Tree (DT) is tree shaped. Each node is a decision
leading to a more precise category of a data for classifi-
cation. It is also an eager learning algorithm because the
tree is constructed during the training process. DTs are
easy to interpret but may produce overfitting.
Random Forest (RF), as its name suggests, uses multiple
DTs in parallel. Each DT is trained on a subset of data and
their results are then merged to determine the most likely
class. This allows reducing the overfitting generated by
DTs. It is an ensemble learning type of algorithm because
it uses multiple other algorithms.
Gradient Boosting (GB) also uses multiple DTs but this
time in a sequence. Each DT learns iteratively on the
errors made by its predecessor. It is also an ensemble
learning kind of algorithm. It can perform better than RF
but potentially has overfitting issues.
E. Evaluation
The performance evaluation of our FDS under the selected
classifiers was done using k-folds cross-validation. This re-
quired splitting the dataset into ksets. k1sets are used
as training and 1 as testing. The process is repeated ktimes
with a different set as the test one. Given that FDSs must be
able to detect falls for new people (e.g. unseen data), the test
set should not contain people data that the algorithm has been
trained on.
We chose a value of k= 5. This creates a training set
of 80% and a test set of 20%. We filtered the SisFall to
only keep subjects that performed all activities. Thus, we
removed all data created by elderly people because they have
not performed simulated falls. We also removed three young
people data due to missing records. This leaves us with
data of 20 subjects that perfectly fit the split in five folds.
Consequently, we have 3400 records consisting of 1900 ADLs
and 1500 falls, making the whole data more balanced.
During the evaluation of ML algorithms, each prediction
falls in one of the following categories:
True Negative (TN): Correct classification of a negative
condition, meaning a reject.
False Positive (FP): Incorrect classification of a negative
condition, meaning a false alarm.
False Negative (FN): Incorrect classification of a positive
condition, meaning a missed.
True Positive (TP): Correct classification of a positive
condition, meaning a hit.
Each prediction is added to the count of its category which
allows then to calculate various metrics such as the accuracy. A
usual representation of these categories is a confusion matrix.
In fall detection, two metrics are especially important:
Sensitivity (SE) (3) and the Specificity (SP) (4) [7]. The SE
(or recall) corresponds to how many relevant elements are
actually selected. This is basically the detection probability
meaning how many falls have actually been detected. The SP
corresponds to how many non-relevant elements are selected.
It means how many non-falls are actually non-falls.
Sensitivity =T P
T P +F N (3)
Specif icity =T N
T N +FP (4)
Additionally, we calculated the accuracy and the Area Under
the Receiver Operating Characteristics Curve (AUROC). The
AUROC is used to evaluate classifiers performance which is
used in pattern recognition and ML [22]. In simple terms,
an AUROC close to the value of one is indicative of a
well-performing algorithm, with high true-positive and true-
negative rates consistently.
IV. RES ULT S AN D DISCUSSION
Table III presents the results of the evaluation of our Fall
Detection System (FDS) under the selected five Machine
Learning (ML) algorithms, showing that we successfully de-
veloped a reliable FDS. Both the Sensitivity (SE) and Speci-
ficity (SP) surpassed 98% with the and the Gradient Boosting
algorithm, outperforming results reported by [21]. In general,
ensemble learning algorithms achieved better performance
than others. This is because they use multiple ML algorithms,
though the improvement in performance is at the expense of
more resources. Support Vector Machine (SVM) had more
difficulties to distinguish the activities. However, by tuning
some hyperparameters, its results would likely improve. In an
FDS, it is desirable to detect every fall to ensure the elderly’s
safety (i.e. a perfect SE). We could then use a threshold to
improve the SE, even though it would reduce the SP and raise
more false alarms.
These results were unexpected especially prior optimization.
We infer that Activities of Daily Living and falls in the SisFall
dataset are discriminating by default, similar to [14]. Thus,
any algorithm can perform very well. However, in real-life
conditions, the SE and SP would very likely drop because
of the falls heterogeneity as highlighted by Krupitzer et al.
[16]. The difficulty of obtaining real falls data is the main
shortcoming in FDS studies, given that it is challenging to
capture them in realistic settings with the elderly.
TABLE III
FALL D ET ECT IO N RES ULTS O F TH E MACHI NE LEARNING ALGORITHMS.
Algorithm Sensitivity
[%]
Specificity
[%]
Accuracy
[%]
AUROC
[%]
KNN 97.26 99.31 98.41 99.45
SVM 87.93 93.78 91.20 96.43
DT 96.60 97.26 96.97 96.93
RF 98.00 98.94 98.52 99.90
GB 98.06 99.21 98.70 99.93
V. CONCLUSIONS AND FU TU RE W OR K
In this paper, we successfully developed a Fall Detection
System (FDS) using wearable technologies. Our results are an
improvement over those reported by [23] and [21], with a final
Sensitivity and Specificity over 98%. The system is reliable
as we were able to test it on a large dataset containing several
thousands of Activities of Daily Living (ADLs) and falls. We
obtained these results using various Machine Learning (ML)
algorithms which we were able to compare. We observed that
ensemble learning algorithms perform better than lazy or eager
learning ones.
There is scope for future work. With the high computa-
tion resources available nowadays, it would be interesting to
explore Deep Learning (DL) algorithms. There is a study
[23] using Recurrent Neural Networks but there are other
algorithms available such as Convolutional Neural Networks
with the advantage of automatic feature extraction from time
series [24]. This reduces the number of steps to implement
and removes the question of how many and which features
are needed to be extracted. The SisFall dataset allows plenty
of experiments. However, the lack of falls data availability in
realistic settings is a common challenge in FDS studies, which
also affected our study.
REFERENCES
[1] L. Z. Rubenstein, “Falls in older people: epidemiology, risk factors
and strategies for prevention,” Age and Ageing, vol. 35, pp. ii37-ii41,
September 2006
[2] World Health Organisation, “WHO global report on falls prevention in
older age,” 2008
[3] S. Sadigh, A. Reimers, R. Andersson, and L. Laflamme, “Falls and
Fall-Related Injuries Among the Elderly: A Survey of Residential-Care
Facilities in a Swedish Municipality,” Journal of Community Health,
vol. 29, pp. 129-140, April 2004
[4] D. Wild, and U. S. Nayak, and B. Isaacs, “How dangerous are falls in
old people at home?,” Br Med J (Clin Res Ed), vol. 282, pp. 266-268,
January 1981
[5] P. Rashidi, and A. Mihailidis, “A Survey on Ambient-Assisted Living
Tools for Older Adults,” IEEE Journal of Biomedical and Health
Informatics, vol. 17, pp. 579-590, May 2013
[6] H. Hawley-Hague, E. Boulton, A. Hall, K. Pfeiffer, and C. Todd, “Older
adults’ perceptions of technologies aimed at falls prevention, detection
or monitoring: A systematic review,” International Journal of Medical
Informatics, vol. 83, pp. 416-426, June 2014
[7] N Noury, A. Fleury, P. Rumeau, A.K. Bourke, G. O. Laighin, V. Rialle,
and J. E. Lundy, “Fall detection - Principles and Methods,” 2007 29th
Annual International Conference of the IEEE Engineering in Medicine
and Biology Society, pp. 1663-1666, August 2007
[8] M. Mubashir, L. Shao, and L. Seed, “A survey on fall detection:
Principles and approaches,” Neurocomputing, vol. 100, pp. 144-152,
January 2013
[9] X. Yu, “Approaches and principles of fall detection for elderly and
patient,” HealthCom 2008 - 10th International Conference on e-health
Networking, Applications and Services, pp. 42-47, July 2008
[10] S. Abbate, M. Avvenuti, F. Bonatesta, G. Cola, P. Corsini, and A.
Vecchio, “A smartphone-based fall detection system,” Pervasive and
Mobile Computing, vol. 8, pp. 883-899, December 2012
[11] A. M. Chan, N. Selvaraj, N. Ferdosi, and R. Narasimhan, “Wireless
patch sensor for remote monitoring of heart rate, respiration, activity,
and falls,” 2013 35th Annual International Conference of the IEEE
Engineering in Medicine and Biology Society (EMBC), pp. 6115-6118,
July 2013
[12] M. Kangas, A. Konttila, P. Lindgren, I. Winblad, and T. J¨
ams¨
a, “Com-
parison of low-complexity fall detection algorithms for body attached
accelerometers,” Gait & Posture, vol. 28, pp. 285-291, August 2008
[13] M. Yuwono, B. D. Moulton, S. W. Su, B. G. Celler, and H. T. Nguyen,
“Unsupervised machine-learning method for improving the performance
of ambulatory fall-detection systems,” BioMedical Engineering OnLine,
vol. 11, pp. 9, February 2012
[14] A. K. Bourke, and G. M. Lyons, “A threshold-based fall-detection
algorithm using a bi-axial gyroscope sensor,” Medical Engineering &
Physics, vol. 30, pp. 84-90, January 2008
[15] J. Wang, Z. Zhang, B. Li, S. Lee, and R. S. Sherratt, “An enhanced fall
detection system for elderly person monitoring using consumer home
networks,” IEEE Transactions on Consumer Electronics, vol. 60, pp.
23-29, February 2014
[16] C. Krupitzer, T. Sztyler, J. Edinger, M. Breitbach, H. Stuckenschmidt,
and C. Becker, “Hips Do Lie! A Position-Aware Mobile Fall Detection
System,” 2018 IEEE International Conference on Pervasive Computing
and Communications (PerCom), pp. 1-10, March 2018
[17] E. Casilari, and M. A. Oviedo-Jim´
enez, “Automatic Fall Detection Sys-
tem Based on the Combined Use of a Smartphone and a Smartwatch,”
PLOS ONE, vol. 10, pp. 1-11, November 2015
[18] T. Vilarinho, B. Farshchian, D. G. Bajer, O. H. Dahl, I. Egge, S. S.
Hegdal, A. Lønes, J. N. Slettevold, and S. M. Weggersen, “A Com-
bined Smartphone and Smartwatch Fall Detection System,” 2015 IEEE
International Conference on Computer and Information Technology;
Ubiquitous Computing and Communications; Dependable, Autonomic
and Secure Computing; Pervasive Intelligence and Computing, pp. 1443-
1448, October 2015
[19] A. T. ¨
Ozdemir, and B. Barshan, “Detecting Falls with Wearable Sensors
Using Machine Learning Techniques,” Sensors, vol. 14, pp. 10691-
10708, June 2014
[20] J. Y. Hwang, J.M. Kang, and Y. W. Jang, and H. C. Kim, “Development
of novel algorithm and real-time monitoring ambulatory system using
Bluetooth module for fall detection in the elderly,” The 26th Annual
International Conference of the IEEE Engineering in Medicine and
Biology Society, pp. 2204-2207, September 2004
[21] A. Sucerquia, J. D. L´
opez, and J. F. Vargas-Bonilla, “SisFall: A Fall and
Movement Dataset,” Sensors, vol. 17, January 2017
[22] T. Fawcett, “An introduction to ROC analysis,” Pattern Recognition
Letters, vol. 27, pp. 861-874, June 2006
[23] M. Musci, D. De Martini, N. Blago, T. Facchinetti, M. and Piastra,
“Online Fall Detection using Recurrent Neural Networks,” arXiv preprint
arXiv:1804.04976, April 2018
[24] F. J. Ord´
o˜
nez, and D. Roggen, “Deep Convolutional and LSTM Recur-
rent Neural Networks for Multimodal Wearable Activity Recognition,”
Sensors, vol. 16, pp. 115, January 2016
... This paper is an extension of our work accepted at the ICAIIC 2020 [8]. This paper has three research questions: RQ1:What is the difference in performance across various types of Machine Learning (ML) algorithms in a FDS? ...
... We then experimented on the filtered data with five ML algorithms. This research question extends our previous work [8]. RQ3:What is the difference in performance across various types of ML algorithms by adopting a multi-class approach for identifying phases of a fall? ...
... These are: the period before the fall happens (pre-fall), the fall itself (impact) and after the fall happened (post-fall). This research question extends our previous work [8]. ...
Article
Full-text available
Falls are dangerous for the elderly, often causing serious injuries especially when the fallen person stays on the ground for a long time without assistance. This paper extends our previous work on the development of a Fall Detection System (FDS) using an inertial measurement unit worn at the waist. Data come from SisFall, a publicly available dataset containing records of Activities of Daily Living and falls. We first applied a preprocessing and a feature extraction stage before using five Machine Learning algorithms, allowing us to compare them. Ensemble learning algorithms such as Random Forest and Gradient Boosting have the best performance, with a Sensitivity and Specificity both close to 99%. Our contribution is: a multi-class classification approach for fall detection combined with a study of the effect of the sensors’ sampling rate on the performance of the FDS. Our multi-class classification approach splits the fall into three phases: pre-fall, impact, post-fall. The extension to a multi-class problem is not trivial and we present a well-performing solution. We experimented sampling rates between 1 and 200 Hz. The results show that, while high sampling rates tend to improve performance, a sampling rate of 50 Hz is generally sufficient for an accurate detection.
... In fall detection, machine learning assists in efficient detection of falls based on patterns of behaviour of the subject. The machine learning classifiers or models for fall detection work upon the datasets collected by either sensor [4][5][6][7][8] or captured with [9][10][11][12] camera. ...
Article
Full-text available
Prior detection for high risk of falls in elderly people is an essential and challenging task. Wearable sensors have already proven as beneficial resource in monitoring daily living activities. Sensors worn on body such as gyroscope, accelerometer can provide a valuable input into detection of fall. In our research, we have implemented the deep learning methods, and analyzed that they are suitable for extracted features from sensors data i.e. accelerometer, gyroscope that evaluate fall risks. We used a publicly available dataset that is based on different daily living activities of elderly people. Furthermore, to conduct the comparative analysis, the performance of two deep learning architectures, the Long short-term memory (LSTM) and CNN based Transfer learning is considered. We also observed that CNN-transfer learning resulted in optimal performance quantitatively bearing 98% accuracy, we summarized that deep learning architectures are very effective in multi-task learning and are capable to effectively predict the high risk of human falls in the terms of wearable sensors.
... The calculated features were used as input to develop a machine learning classifier for pre-impact fall detection based on an adaptive boosting algorithm [41]. Gradient boosting algorithms, such as AdaBoost, have been noted for superior fall detection performance as compared to other methods including k-Nearest Neighbor (KNN), Support Vector Machine (SVM), and Random Forest (RF) [42]. The AdaBoost classifier is a multiple decision tree classifier that, in sequence, chooses a selective weight by which the average contributes to the final prediction of the test data set. ...
Article
Full-text available
Background: Falls are a common complication experienced after a stroke and can cause serious detriments to physical health and social mobility, necessitating a dire need for intervention. Among recent advancements, wearable airbag technology has been designed to detect and mitigate fall impact. However, these devices have not been designed nor validated for the stroke population and thus, may inadequately detect falls in individuals with stroke-related motor impairments. To address this gap, we investigated whether population-specific training data and modeling parameters are required to pre-detect falls in a chronic stroke population. Methods: We collected data from a wearable airbag's inertial measurement units (IMUs) from individuals with (n = 20 stroke) and without (n = 15 control) history of stroke while performing a series of falls (842 falls total) and non-falls (961 non-falls total) in a laboratory setting. A leave-one-subject-out crossvalidation was used to compare the performance of two identical machine learned models (adaptive boosting classifier) trained on cohort-dependent data (control or stroke) to pre-detect falls in the stroke cohort. Results: The average performance of the model trained on stroke data (recall = 0.905, precision = 0.900) had statistically significantly better recall (P = 0.0035) than the model trained on control data (recall = 0.800, precision = 0.944), while precision was not statistically significantly different. Stratifying models trained on specific fall types revealed differences in pre-detecting anterior-posterior (AP) falls (stroke-trained model's F1-score was 35% higher, P = 0.019). Using activities of daily living as non-falls training data (compared to near-falls) significantly increased the AUC (Area under the receiver operating characteristic) for classifying AP falls for both models (P < 0.04). Preliminary analysis suggests that users with more severe stroke impairments benefit further from a stroke-trained model. The optimal lead time (time interval pre-impact to detect falls) differed between control- and stroke-trained models. Conclusions: These results demonstrate the importance of population sensitivity, non-falls data, and optimal lead time for machine learned pre-impact fall detection specific to stroke. Existing fall mitigation technologies should be challenged to include data of neurologically impaired individuals in model development to adequately detect falls in other high fall risk populations. Trial registration https://clinicaltrials.gov/ct2/show/NCT05076565 ; Unique Identifier: NCT05076565. Retrospectively registered on 13 October 2021.
... To train the classifiers and perform the classification, a number of features have been extracted from the datasets. In this study, eight time-domain features and one frequencydomain feature [52] were extracted from each accelerometer channel as well as from the absolute summation (s(t)) of the 3D time series data (a(t) = [a x (t), a y (t), a z (t)]). The features were mean, median, standard deviation, variance, min, max, 0.25 quantile, 0.75 quantile, and spectral entropy. ...
Article
Full-text available
Falls are a major cause of morbidity and mortality in neurological disorders. Technical means of detecting falls are of high interest as they enable rapid notification of caregivers and emergency services. Such approaches must reliably differentiate between normal daily activities and fall events. A promising technique might be based on the classification of movements based on accelerometer signals by machine-learning algorithms, but the generalizability of classifiers trained on laboratory data to real-world datasets is a common issue. Here, three machine-learning algorithms including Support Vector Machine (SVM), k-Nearest Neighbors (kNN), and Random Forest (RF) were trained to detect fall events. We used a dataset containing intentional falls (SisFall) to train the classifier and validated the approach on a different dataset which included real-world accidental fall events of elderly people (FARSEEING). The results suggested that the linear SVM was the most suitable classifier in this cross-dataset validation approach and reliably distinguished a fall event from normal everyday activity at an accuracy of 93% and similarly high sensitivity and specificity. Thus, classifiers based on linear SVM might be useful for automatic fall detection in real-world applications.
Conference Paper
With the increasing global aging population, the health of the elderly has become a global concern. Accidental falls, as one of the major causes of health and safety issues affecting the elderly, can cause serious hazards. In this paper, a fall detection system is proposed to be able to deliver timely information after a fall. The acceleration and angular velocity time series extracted from motion were used to describe human motion features. Hybrid threshold analysis algorithm and machine learning algorithm are used for classification between falls and activities of daily living (ADLs). The fall detection results showed 98.55% accuracy, 98.16% sensitivity, and 98.73% specificity. The result is higher than the single-threshold algorithm and slightly lower than the machine learning algorithm. In addition, the hybrid algorithm of fall detection in this paper is to put the threshold analysis algorithm in the edge device for calculation and put the machine learning algorithm in the cloud server for calculation. Since the single machine learning algorithm needs to transmit data to the cloud server all the time, the hybrid algorithm has lower power consumption than machine learning algorithms, and the average alarm time is shorter, making it more suitable for actual systems.
Article
Full-text available
Telemonitoring of human physiological data helps detect emergency occurrences for subsequent medical diagnosis in daily living environments. One of the fatal emergencies in falling incidents. The goal of this paper is to detect significant incidents such as falls. The fall detection system is essential for human body movement investigation for medical practitioners, researchers, and healthcare businesses. Accelerometers have been presented as a practical, low-cost, and dependable approach for detecting and predicting outpatient movements in the user. The accurate detection of body movements based on accelerometer data enables the creation of more dependable systems for incorporating long-term development in physiological remarks. This research describes an accelerometer-based platform for detecting users' body movement when they fall. The ADXL345, MMA8451q, and ITG3200 body sensors capture activity data, subsequently classified into 15 fall incident classes based on SisFall dataset. Falling incidents classification is performed using Long Short-Term Memory results in best AUC-ROC value of 97.7% and best calculation time of 6.16 seconds. Meanwhile, Support Vector Machines results in the best AUC-ROC value of 98.5% and best calculation times of 17.05 seconds.
Article
Full-text available
This work implements a fall and posture detection system exploiting low cost sensors and applying machine learning to aid people in need both at indoor and outdoor. This intelligent system is able to identify fall with and without recovery within a stipulated period of time. In case of fall without recovery, an alert message along with date, time and location of fall is sent to relative/caregiver. This feature ensures real time assistance to avoid any criticality due to delay. In addition to this, an immediate last posture before the fall is also notified to identify the proneness of a person towards fall from a specific posture. This may aid clinical persons to take appropriate measures to prevent the future fall. The system is also able to take care of an unresponsive device after a fall (if any). We have designed and implemented this intelligent live fall with posture detection system, exploiting the sensors in micro processor unit (MPU) 6050 combined with low cost ESP 8266 micro-controller unit (MCU) using WiFi connectivity. The kinematic sensor data is collected at a rate of 40 Hz using accelerometer and gyroscope.The result shows that the system can identify the location and posture of the subject on regular interval along with the date and time of fall (if any). The emergency help system is aided with an audio-visual warning at the raspberry Pi based monitoring station along with a facility of sending the distress SMS.The system can operate either in manual or in auto mode. The dataset is prepared from local people of varied age groups (between 10 and 70 years) of both the genders.The system is tested randomly on 10 volunteers with an overall detection accuracy upto 98%.
Article
Full-text available
Unintentional falls can cause severe injuries and even death, especially if no immediate assistance is given. The aim of Fall Detection Systems (FDSs) is to detect an occurring fall. This information can be used to trigger the necessary assistance in case of injury. This can be done by using either ambient-based sensors, e.g. cameras, or wearable devices. The aim of this work is to study the technical aspects of FDSs based on wearable devices and artificial intelligence techniques, in particular Deep Learning (DL), to implement an effective algorithm for on-line fall detection. The proposed classifier is based on a Recurrent Neural Network (RNN) model with underlying Long Short-Term Memory (LSTM) blocks. The method is tested on the publicly available SisFall dataset, with extended annotation, and compared with the results obtained by the SisFall authors.
Article
Full-text available
Research on fall and movement detection with wearable devices has witnessed promising growth. However, there are few publicly available datasets, all recorded with smartphones, which are insufficient for testing new proposals due to their absence of objective population, lack of performed activities, and limited information. Here, we present a dataset of falls and activities of daily living (ADLs) acquired with a self-developed device composed of two types of accelerometer and one gyroscope. It consists of 19 ADLs and 15 fall types performed by 23 young adults, 15 ADL types performed by 14 healthy and independent participants over 62 years old, and data from one participant of 60 years old that performed all ADLs and falls. These activities were selected based on a survey and a literature analysis. We test the dataset with widely used feature extraction and a simple to implement threshold based classification, achieving up to 96% of accuracy in fall detection. An individual activity analysis demonstrates that most errors coincide in a few number of activities where new approaches could be focused. Finally, validation tests with elderly people significantly reduced the fall detection performance of the tested features. This validates findings of other authors and encourages developing new strategies with this new dataset as the benchmark.
Article
Full-text available
Human activity recognition (HAR) tasks have traditionally been solved using engineered features obtained by heuristic processes. Current research suggests that deep convolutional neural networks are suited to automate feature extraction from raw sensor inputs. However, human activities are made of complex sequences of motor movements, and capturing this temporal dynamics is fundamental for successful HAR. Based on the recent success of recurrent neural networks for time series domains, we propose a generic deep framework for activity recognition based on convolutional and LSTM recurrent units, which: (i) is suitable for multimodal wearable sensors; (ii) can perform sensor fusion naturally; (iii) does not require expert knowledge in designing features; and (iv) explicitly models the temporal dynamics of feature activations. We evaluate our framework on two datasets, one of which has been used in a public activity recognition challenge. Our results show that our framework outperforms competing deep non-recurrent networks on the challenge dataset by 4% on average; outperforming some of the previous reported results by up to 9%. Our results show that the framework can be applied to homogeneous sensor modalities, but can also fuse multimodal sensors to improve performance. We characterise key architectural hyperparameters’ influence on performance to provide insights about their optimisation.
Conference Paper
Full-text available
Falls are behind many elderly hospitalizations and can lead to injuries that greatly debilitate old patients. Much of the deployed fall detection systems rely on the user wearing a personal emergency response device, being conscious and at home. The limitations of the existing systems regarding usability and efficiency have yield an overarching research question on whether systems based on new and advanced consumer mobile devices can be used as ubiquitous automatic fall detectors for seniors. This paper specifically looks into the accuracy of a fall detection system based on an off-the-shelf smartwatch and smartphone. We have implemented a system which combines threshold based and pattern recognition techniques in both devices, with the intent of having the watch to contribute to the specificity of the fall detection strategy. We tested the accuracy of the system through a series of simulated falls and activities of daily living, resulting on the correct identification of 63% of the falls and 78% of the activities and outperforming two other baseline fall detection applications (iFall and Fade). The sensors and algorithm on the watch were able to provide a marginal contribution to the system’s accuracy. Indications from the tests suggest that it should be possible to improve the system accuracy by adjusting the used thresholds and fuzzyfying them. Moreover, it is expected that the open source nature of this work and it’s results boost such threshold tuning and serve as a better basis for researchers to benchmark their work. Accepted version at: https://www.tvilarinho.com/
Article
Full-text available
Due to their widespread popularity, decreasing costs, built-in sensors, computing power and communication capabilities, Android-based personal devices are being seen as an appealing technology for the deployment of wearable fall detection systems. In contrast with previous solutions in the existing literature, which are based on the performance of a single element (a smartphone), this paper proposes and evaluates a fall detection system that benefits from the detection performed by two popular personal devices: a smartphone and a smartwatch (both provided with an embedded accelerometer and a gyroscope). In the proposed architecture, a specific application in each component permanently tracks and analyses the patient’s movements. Diverse fall detection algorithms (commonly employed in the literature) were implemented in the developed Android apps to discriminate falls from the conventional activities of daily living of the patient. As a novelty, a fall is only assumed to have occurred if it is simultaneously and independently detected by the two Android devices (which can interact via Bluetooth communication). The system was systematically evaluated in an experimental testbed with actual test subjects simulating a set of falls and conventional movements associated with activities of daily living. The tests were repeated by varying the detection algorithm as well as the pre-defined mobility patterns executed by the subjects (i.e., the typology of the falls and non-fall movements). The proposed system was compared with the cases where only one device (the smartphone or the smartwatch) is considered to recognize and discriminate the falls. The obtained results show that the joint use of the two detection devices clearly increases the system’s capability to avoid false alarms or ‘false positives’ (those conventional movements misidentified as falls) while maintaining the effectiveness of the detection decisions (that is to say, without increasing the ratio of ‘false negatives’ or actual falls that remain undetected).
Article
Full-text available
Falls are a serious public health problem and possibly life threatening for people in fall risk groups. We develop an automated fall detection system with wearable motion sensor units fitted to the subjects' body at six different positions. Each unit comprises three tri-axial devices (accelerometer, gyroscope, and magnetometer/compass). Fourteen volunteers perform a standardized set of movements including 20 voluntary falls and 16 activities of daily living (ADLs), resulting in a large dataset with 2520 trials. To reduce the computational complexity of training and testing the classifiers, we focus on the raw data for each sensor in a 4 s time window around the point of peak total acceleration of the waist sensor, and then perform feature extraction and reduction. Most earlier studies on fall detection employ rule-based approaches that rely on simple thresholding of the sensor outputs. We successfully distinguish falls from ADLs using six machine learning techniques (classifiers): the k-nearest neighbor (k-NN) classifier, least squares method (LSM), support vector machines (SVM), Bayesian decision making (BDM), dynamic time warping (DTW), and artificial neural networks (ANNs). We compare the performance and the computational complexity of the classifiers and achieve the best results with the k-NN classifier and LSM, with sensitivity, specificity, and accuracy all above 99%. These classifiers also have acceptable computational requirements for training and testing. Our approach would be applicable in real-world scenarios where data records of indeterminate length, containing multiple activities in sequence, are recorded.
Article
Full-text available
Various fall-detection solutions have been previously proposed to create a reliable surveillance system for elderly people with high requirements on accuracy, sensitivity and specificity. In this paper, an enhanced fall detection system is proposed for elderly person monitoring that is based on smart sensors worn on the body and operating through consumer home networks. With treble thresholds, accidental falls can be detected in the home healthcare environment. By utilizing information gathered from an accelerometer, cardiotachometer and smart sensors, the impacts of falls can be logged and distinguished from normal daily activities. The proposed system has been deployed in a prototype system as detailed in this paper. From a test group of 30 healthy participants, it was found that the proposed fall detection system can achieve a high detection accuracy of 97.5%, while the sensitivity and specificity are 96.8% and 98.1% respectively. Therefore, this system can reliably be developed and deployed into a consumer product for use as an elderly person monitoring device with high accuracy and a low false positive rate.
Article
Full-text available
In recent years, we have witnessed a rapid surge in assisted living technologies due to a rapidly aging society. The aging population, the increasing cost of formal health care, the caregiver burden, and the importance that the individuals place on living independently, all motivate development of innovative-assisted living technologies for safe and independent aging. In this survey, we will summarize the emergence of 'ambient-assisted living" (AAL) tools for older adults based on ambient intelligence paradigm. We will summarize the state-of-the-art AAL technologies, tools, and techniques, and we will look at current and future challenges.
Article
Over recent years a number of Information and Communication Technologies (ICTs) have emerged aiming at falls prevention, falls detection and alarms for use in case of fall. There are also a range of ICT interventions, which have been created or adapted to be pro-active in preventing falls, such as those which provide strength and balance training to older adults in the prevention of falls. However, there are issues related to the adoption and continued use of these technologies by older adults. This review provides an overview of older adults' perceptions of falls technologies. We undertook systematic searches of MEDLINE, EMBASE, CINAHL and PsychINFO, COMPENDEX and the Cochrane database. Key search terms included 'older adults', 'seniors', 'preference', 'attitudes' and a wide range of technologies, they also included the key word 'fall*'. We considered all studies that included older adults aged 50 and above. Studies had to include technologies related specifically to falls prevention, detection or monitoring. The Joanna Briggs Institute (JBI) tool and the Quality Assessment Tool for Quantitative Studies by the Effective Public Health Practice Project (EPHPP) were used. We identified 76 potentially relevant papers. Some 21 studies were considered for quality review. Twelve qualitative studies, three quantitative studies and 6 mixed methods studies were included. The literature related to technologies aimed at predicting, monitoring and preventing falls suggest that intrinsic factors related to older adults' attitudes around control, independence and perceived need/requirements for safety are important for their motivation to use and continue using technologies. Extrinsic factors such as usability, feedback gained and costs are important elements which support these attitudes and perceptions. Positive messages about the benefits of falls technologies for promoting healthy active ageing and independence are critical, as is ensuring that the technologies are simple, reliable and effective and tailored to individual need. The technologies need to be clearly described in research and older peoples' attitudes towards different sorts of technologies must be clarified if specific recommendations are to be made.