=Paper=
{{Paper
|id=Vol-2996/paper4
|storemode=property
|title=Personal Hygiene Monitoring Under the Shower Using WiFi Channel State Information
|pdfUrl=https://ceur-ws.org/Vol-2996/paper4.pdf
|volume=Vol-2996
|authors=Jeroen Klein Brinke,Alessandro Chiumento,Paul Havinga
|dblpUrl=https://dblp.org/rec/conf/ewsn/BrinkeCH21
}}
==Personal Hygiene Monitoring Under the Shower Using WiFi Channel State Information==
https://ceur-ws.org/Vol-2996/paper4.pdf
Personal Hygiene Monitoring Under the Shower Using
Wi-Fi Channel State Information
Jeroen Klein Brinke Alessandro Chiumento Paul Havinga
j.kleinbrinke@utwente.nl a.chiumento@utwente.nl p.j.m.havinga@utwente.nl
University of Twente University of Twente University of Twente
Enschede, Netherlands Enschede, Netherlands Enschede, Netherlands
ABSTRACT technologies [17], wearable technologies [9], or infrastructural sen-
Personal hygiene is often used to measure functional independence, sors (e.g. pressure mats or door sensors) [2, 4, 18]. The audiovisual
which is how much support someone requires to perform self-care. technologies cannot be placed in the shower as these are privacy-
By extension, this is often used in the monitoring of (early-stage) invasive. When it comes to wearable sensors, they may need to
dementia. Current technologies are based on either audiovisual or be taken off during shower time [9], unless they are waterproof.
wearable technologies, both of which have practical limitations. However, it cannot be assumed that elderly people and patients
The use of (NLOS) radio-frequency based human activity recogni- suffering from dementia never forget to wear a wearable device
tion could provide solutions here. This paper leverages the 802.11n or know how to use properly use them. Lastly, infrastructural sen-
channel state information to monitor different shower-related ac- sors often require great modification to existing homes (such as
tivities (e.g. washing head or body, brushing teeth, and dressing installing sensors in the ceiling, walls, and door frames) and these
up) and the degree to which some of these can be monitored, as sensors are often bound to a specific location, whether that is a
well estimating different water pressures used while showering for room or an object (such as a door frame).
multiple locations in the apartment. Wavelet denoising is applied Unlike the aforementioned technologies, radio frequency-based
for filtering and a convolutional neural network is implemented for technologies (unobtrusive sensing) do not have to be put inside
classification. Results imply that for coarse-grained activity recog- the shower itself, are more robust (e.g. do not need to be worn
nition, an πΉ 1 -score of 0.85 is achievable for certain classes, while for and cannot be forgotten) and are likely more privacy-aware, as the
fine-grained this drops to 0.75. Water pressure estimation ranges data is not easily interpretable by humans, since it requires more
from 0.75 to 0.85 between fine-grained and coarse-grained, respec- complex (pre)processing. This enables ADL monitoring [3, 16],
tively. Overall, this paper shows that channel state information can abnormal activity detection [24], and vital sign monitoring [13, 21]
be successfully employed to monitor variations in different shower using RF-based technologies. An activity often considered in health
activities, as well as successfully estimating the water pressure in care, and even more so in elderly and dementia care, is falling
the shower. [10, 14, 19, 22]. Falling is dangerous and happens most frequently
in patientβs rooms and bathrooms [11], with the actual room being
CCS CONCEPTS ahead of the bathroom. However, falling in the bathroom often
results in more serious injuries.
β’ Computer systems organization β Sensor networks; β’ Net-
This paper leverages channel state information in a non line-
works β Wireless local area networks; β’ Human-centered com-
of-sight (NLOS) environment to monitor the degree of shower be-
puting β Ubiquitous and mobile computing.
haviour on multiple locations in an apartment. This degree comes
in the form of several shower-related activities, but performed in
KEYWORDS two different ways: one regular showering (as a person would), the
channel state information, human activity recognition, device-free other is by acting as if showering is troublesome (slow, painful,
sensing, 802.11n, personal hygiene, deep learning troublesome). Additionally, the activities of (un)dressing, drying of
the body and brushing of the teeth are considered in their regular
1 INTRODUCTION form. Results suggest that RF-based sensing can be used to inves-
tigate the aforementioned activities with an πΉ 1 -score score over
Personal hygiene is an important aspect in elderly care and even 0.85, while also achieving a comparable πΉ 1 -score score for water
more so for those suffering from (early-stage) dementia. It is an pressure estimation, depending of the location of the receiver.
important determinant of the level of support someone requires to This paper first discusses the related works, which include works
perform self-care, which is part of the functional independence in which inspired this research, as well as current solutions using other
activities of daily living (ADL) [1, 5, 6, 8, 15]. Most accepted solu- technologies. After that the data acquisition is discussed, as well
tions to monitor bathroom activities focus on the use of audiovisual as an overview of the dataset, to encourage fellow researcher to
replicate this work and contribute to the presented dataset. During
Copyright 2021 for this paper by its authors. Use permitted under Creative Commons
License Attribution 4.0 International (CC BY 4.0).
data acquisition, the variables are also discussed and how these are
chosen. Following this, the actual methodology is discussed, which
includes the ground truth evaluation, signal processing and data
analysis. In the following Results and Discussion section, the actual
outcome is evaluated and the research questions will be answered.
CHIIoT 1, February 17, 2021, Delft, The Netherlands Klein Brinke, et al.
Finally, the paper concludes with a short summary of the results water pressures and different gradations of performing the shower
and suggestions for future work. activities.
Zhang et al. [23] considered different poses in a bath using a
1.1 Challenges and contribution single transmitter and receiver pair on the 5 GHz band: one regular
Currently, little to no research has gone into personal hygiene (steady lying position) and two dangerous poses (the whole body
detection under the shower, which is used in the monitoring of sunk to the bottom and face-down in a bath). Data was collected by
(early-stage) dementia. The use of audiovisual, wearable of infras- a single volunteer over the course of multiple months in a single
tructural monitoring technologies has been proven [2, 4, 9, 17, 18], apartment. Results show that an πΉ 1 -score of 89.47% can be achieved
but come with downsides in privacy, installation and robustness. using this system. Zhang et al. considers the drowning dangers in a
Most existing RF-based solutions focus either on detecting coarse- bath tub, but houses and especially smaller apartments often do not
grained shower activities (such as washing or brushing teeth) and come with bath tubs. Therefore, this paper considers the shower to
not on the degree to which these activities are performed. be another important source for personal hygiene and investigates
The use of indirect (NLOS) radio-frequency based sensing could the most prominent activities.
change this, as it reduced both the immediate privacy issues (video Wang et al. [20] developed E-eyes on the 5 GHz band with three
cameras under the shower) and the need for water-proof wearable off-the-shelf Wi-Fi devices connected to a single access point. The
devices and the wireless communication. Additionally, no infras- experiments are conducted in two apartments and the activities
tructural changes to a home are required, such as expensive shower are performed by four male adults. Nine daily activities are con-
heads, or tiles- and wall-mounted sensors. Due to the unobtrusive sidered, including bathing and brushing teeth, which were both
nature, RF-based sensing could prove to be useful in monitoring performed in the bathroom. Wang et al. remark that the two activi-
the variation to which shower activities are being performed, as ties have small differences in their CSI patterns [20, p. 8]. However,
well as combine these with shower information (e.g. water pres- using wider-band signals, the false positive rate is lower than 1%.
sure). Multiple questions arise here in relation to the positioning While this research does consider different (smaller) activities in
of different receivers, but these will be discussed during the data the shower, it does not consider different receiver placement, water
acquisition and methodology. pressures, nor different degrees of the same activity.
The contributions of this paper are:
β’ To show the extend to which variations of different activities
can be identified and monitored under the shower
β’ To show the extend to which different water pressures can
be identified and monitored under the shower
β’ To investigate the effect of different receiver placement on
the accuracy of the aforementioned aspects
2 RELATED WORKS
Channel state information is one of the most prominent RF-based
sensing techniques. It gathers information regarding the signal mul-
tipath propagation between a transmitter and receiving antenna at
the receiver side. Multipath propagation is a phenomenon caused
by environmental influences on the signal, such as scattering, ab-
sorbing, and reflecting. These environmental influences include
humans, and when monitoring the changes in the channel state
information over time, the activities of them. Channel state infor-
mation contains information on the phase and amplitude of the
received signal. The combination of all these antenna pairs is col-
lected in a channel state information matrix, which has the shape
of ππ β ππ‘ β ππ , where ππ is the number of receiving antennas,
ππ‘ the number of transmitting antennas, and ππ the number of
subcarriers.
Research in activity recognition under the shower is fairly lim-
ited. Lee et al. [12] developed a system based on one transmitter
Figure 1: Apartment layout for the WiSh experiments.
and multiple receivers for activity recognition using 802.11n chan-
Green triangles represent the receivers, the red triangle rep-
nel state information. As part of the research, different activities
resents the transmitter. The area in blue is the actual bath-
were considered, including bathing and toileting. It is shown that
room. Letters indicate both the location, as well as the iden-
these activities can be detected with a high accuracy (higher than
tifier of the receiver, where LR=Living Room, FB=Fuse box,
95%) in two test beds when combining the receivers. However, fine-
SR=Study Room, BR=Bed Room
grained shower activity recognition and water pressure estimation
are omitted. This paper will provide a deeper insight into different
Personal hygiene monitoring under the shower using Wi-Fi channel state information CHIIoT 1, February 17, 2021, Delft, The Netherlands
3 DATA ACQUISITION Table 1: List of all experimental parameters, including the
activity abbreviations, where shower-related activities are
3.1 Experimental setup in bold
The setup consisted of a multiple custom Gigabyte Brix IoT devices,
each consisting of an Intel Apollo Lake N34500 processor, 8GB Parameter Values Count
DDRL 1866 MHz memory, and a n Intel N Ultimate Wi-Fi Link 5300 Participants 0,1 2
as a network interface. For these experiments, one functioned as a Activities Undress (UD), dry 9
transmitter and the others as receivers. While the Linux CSI Tool body (DB), idle (I),
[7] offers multiple options for connectivity, ultimately the injection wash hair and face
mode was used. As for the frequency band, 5 GHz was used, as it excitedly (WHF-E),
is better capable in activity fine-grained activities due to its short wash body and legs
wavelength (π = 6.0ππ). excitedly (WBL-E),
While multiple locations were considered for the transmitter, wash hair and face
ultimately a setup was picked that consists of four receivers and a slowly (WHF-S),
single transmitter (Figure 1). Here, the transmitter (red triangle) is wash body and legs
located in the farthest corner from the apartment, where the actual slowly (WBL-S),
Wi-Fi router would be located in this real-life setting. The first brush teeth (BT), dress
receiver is positioned in direct line-of-sight from the transmitter up (DU)
(βΌ 7π) as a zero-measurement (LR). The second receiver is placed Water pressures Off, low, med(ium), 4
in the fuse box compartment of the apartment (FB), positioning it max(imum)
skewed behind the shower (βΌ 9.6π), as it is close to the shower and Receivers LR,FB,SR,BR 4
the potential water pipes leading into it. The third receiver is located Total #combinations 288
in the study room (SR), directly behind the the bath room. However,
while the receiver is theoretically closer than receiver FB (βΌ 8.5π),
the signal needs to propagate through more obstacles (e.g. walls suffering from (early-stage) dementia. Therefore, some activities
and doors). The fourth and final receiver is placed in the bed (BR), are not necessarily bound to showering (e.g. washing head/face),
which would theoretically result in the highest coverage (ignoring but also to other activities related to personal hygiene (e.g. brushing
any obstruction). While it has the best propagation conditions (least teeth).
overlap with the shower), it is the farthest away form the transmitter All activities are considered with four different water pressures:
(13.9π). off, low, medium, and high. It should be noted that these pressures
It should be noted that the data was collected over multiple days, can be minimized into a binary problem, namely off and on (low,
with the apartment being lived in regularly. This means furniture medium, and high combined). It is likely that the water pressure is a
may be moved around and there could be slight changes to the personal preference, rather than there being a correlation between
position of the receivers. Overall, this setup is more relevant than a personal hygiene and the water pressure. Therefore, the binary
fixed laboratory setup, as it represents the actual use-case of such a problem of the shower being on or off is most prominent.
system in the future. However, it was ensured no additional faucets The activities under the shower are performed in two manners:
(e.g. kitchen sink or dishwasher) were on while showering, as this one in a regular/excited fashion, the other in a slow/demotivated
was out of the scope for this research. fashion. This is done to see if a differentiation can be made between
It is assumed that participants likely live alone or shower at either, in order to make it possible to monitor the degradation of
times they are alone. However, it should be noted that there were personal hygiene in patients with (early-stage) dementia. For regu-
two people in the apartment at all times during these experiments: lar/excited, it is suggested the participants shower as they normally
the person not showering was always sitting at the kitchen table would, or in a very good mood. This could be different depending
as far away as possible, minimizing the amount of movement on on the participant, as the definition of a very good mood and the
the signals. Cross-activity recognition among multiple people (e.g. results of such a mood differ per participant. For slow/demotivated,
one shower, one cooking) is not considered as part of this paper. participants are asked to pretend like they are either physically
Therefore, the the receiver in the living room is used to verify restricted or in a very bad mood while showering.
this has a minimum influence on the data collected by the other Due to the sensitive of the data, no ground truth could be col-
receivers. While concrete details on the participants cannot be given lected. Rather, participants are asked to follow a set of instructions
due to privacy concerns, it can disclosed that the two participants (playing through a speaker in the shower) in order to have anal-
were a male and female, with a height difference of 15cm and a ogous activities in the same time slots. Between each activity is
weight difference of 20kg. This shows the participantsβ physiques a moment for idling, which is used for both classification and to
were not similar. separate different activities.
No smart shower head was used, so the exact L/min is unknown
3.2 Activities and water pressures and it is likely that per run, there is a slight difference between each.
For this research, 9 activities are considered (Table 1). These ac- However, an attempt was made to replicate each shower based on
tivities are all based inside the bathroom, e.g. activities occurring the visual appearance and sound of the water beam coming out of
under or around the shower in healthy participants and patients the shower head. For Low, the shower should drizzle: barely any
CHIIoT 1, February 17, 2021, Delft, The Netherlands Klein Brinke, et al.
water should come through the shower head. Med is the setting at 4.2.3 Receiver positioning. Finally, the activity classification and
which there is just enough pressure in the shower head to result in the water pressure estimation are compared between the different
a consistent stream. The last setting, Max, requires the shower to receivers (Figure 1). The most optimal location for these experi-
be on at full force. ments will be discussed, as well as the worst performing one. For
this, only the shower-related activities are used for both activity
4 METHODOLOGY and water pressure recognition. Receiver LR is omitted, as it was
used as a zero-measurement (no major movements happening in
4.1 Prepocessing the living room).
For preprocessing, the first step is restructuring of the channel state
information, which is a 4D-matrix (3 β 3 β 30 β π‘) due to 3x3 MIMO
and 30 subcarriers over time π‘, into the shape of the input layer
to the convolutional neural network, which has a shape of a 3D-
matrix (π» β π β π·). For these experiments, the data was flattened
into a 3D-shape with a depth of 1 (π· = 1), namely 270 β π‘ β 1,
as 3 β 3 β 30 = 270. This means that for every antenna pair, the
subcarriers are stacked on top of each other (e.g. the first 30 rows
are receiving antenna 1 and transmitting antenna 1, the second 30
rows are receiving antenna 1 and transmitting antenna 2, and so
on). Afterwards, wavelet denoising was used to denoise the signal
in MATLAB, after which the data was stored as a set of images,
where each pixels is thus based on an absolutely value for a specific
subcarrier in a specific antenna-pair (π» ) at time π‘ (π ).
4.2 Classification
A convolutional neural network (CNN) is employed to for classi-
fication, as it preserves the spatial and structural information of
the channel state information. The CNN consists of three 2D con-
volution layers, with a 0.6-dropout after the first and third layer. (a)
Max-pooling, batch normalization and a leaky ReLU (πΌ = 0.1) for
the activation layers are applied after each layer. At the end, the
Figure 2: Classification of activities based on 10 runs with
outputs are flattened and go through a 160-neuron dense layer with
250 epochs for π = {0, 1}, π = ππ
, which includes all activities
the sigmoid activation, before reaching the final dense layer for
as mentioned in Table 1
classification (softmax). The model was trained for 250 epochs, with
a batch size of 8. The learning rate started with an initial learning
rate of 1 β 10β4 , with a decay rate of 0.95 every 50 steps. The split
between the training and testing set used is 0.60 and 0.40 for the 5 RESULTS AND DISCUSSION
data of both users combined, respectively. 5.1 Classification
4.2.1 Activity classification. The first thing to consider is the ac- 5.1.1 Activity classification. Figure 2 presents the classification as a
tivity classification in, regardless of the water pressure. For this, confusion matrix for all activities for π = ππ
, π = 0, as all activities
the shower-related activities (bold in Table 1) are combined over are important for the activity recognition part. While Figure 2
the different water pressures (including Off ). This is to give an shows the normalized confusion matrix over 10 runs, the average
indication whether or not activity classification is affected by water πΉ 1 -score with standard deviation over these 10 runs are used as a
pressure. A differentiation is made between all all activities and metric when discussing specific classes. The overall πΉ 1 -score for
shower-related activities, where all activities include activities only all activities is 0.74 Β± 0.05. For the case of all activities, 4 out of 9
happening during Off and where shower-related activities only classes have an πΉ 1 -score over 0.85 (BT, DU, DB, WHF-E) and 3 more
include those while the shower is on. This is evaluated over the an accuracy over 0.75 (WHF-S, WBL-E/S). The only exceptions here
different receiver locations. are idle (I ) 0.4 Β± 0.0 and undress (UD), with an πΉ 1 -score of 0.58 Β± 14
and 0.03 Β± 0.09, respectively.
4.2.2 Water pressure estimation. Here the labeling is based on the For undress, this is likely due to the limited data, as the activity
water pressure. All data is combined over the different activities only lasts for 30 seconds. This causes only 40 data fragments after
depending on the water pressure. At the lowest level, the binary preprocessing. As a comparison, all other classes have at least 110
problem of shower On/Off is considered by clustering everything data fragments, with the showering activities (WHF-E/S,WBL-E/S,I )
other than Off as On. Afterwards, this is turned into a more fine- over 330. Additionally, undressing takes place at approximately the
grained problem where an attempt is made to differentiate per same location as brushing teeth.
individual water pressure. For the classification of water pressure, For Idle it is a bit different, as it has a comparable number of
only the shower-related activities are considered. data frames after preprocessing, namely 334 data frames. Idle is
Personal hygiene monitoring under the shower using Wi-Fi channel state information CHIIoT 1, February 17, 2021, Delft, The Netherlands
mostly confused with brushing teeth (0.13 Β± 0.01) and washing of between the two and Med, as can be observed in the confusion
the both the head and body (in the range of 0.11) for both slow matrices by the darker colors and upon further inspection of the
and excited, likely due to these being activities involving minor actual classification rates. This indicates a larger difference between
body movements. Minor body movements are a part of idling, as Low and Max, but lesser so between the two and Med, which can
participants cannot be expected to stand completely still in the be explained by small differences in the shower sessions and the
shower (e.g. getting water in the eye or uncomfortable positioning). variation in the shower head: sometimes, medium may be exactly
Another observation is that there are darker squares in the con- between low and maximum, but other times it may edge more
fusion matrix around the two different performances (excited and towards either of the two.
slow) of Washing body and legs and Washing hair and face. For
Washing body and legs and Washing hair and face, the πΉ 1 -score is 5.1.3 Receiver positioning. Figure 4 shows the additional classi-
0.77Β±0.15β0.81Β±0.07 and 0.80Β±0.10β0.87Β±0.07, respectively. The fication performance for receivers FB (a,c), and BR (b,d). The top
false-positives and true-negatives are in the range of 0.15, which is and bottom rows show the classification accuracy over 250 runs for
likely due to it being the same activity being performed with varia- π = {0, 1} for activities and water pressure recognition, respectively.
tion: sometimes an optimistic interpretation of slowly washing is The classification is based on the relevant activities, as these are
close to a pessimistic interpretation of excited washing. Overall, this the best case scenarios. The classification for all activities performs
implies that an estimation can be made regarding the performance worse, as is previously discussed in sections 5.1.1 and 5.1.2. The
of washing, while there is a clearer distinction between washing information for receiver SR can be found in Figure 2 and 3 for
the body and legs and washing the face. It is likely that the accu- activities and water pressure, respectively.
racy will drop when even more fine-grained activity recognition is It can be seen that receiver SR is the most prominent at identify-
considered (e.g. washing hair, face, upper body, and lower body). ing the activities (πΉ 1 -score of 0.74 Β± 0.05) , while receiver BR is most
prominent at water pressure estimation (0.88 Β± 0.09). While both
score comparable on activity recognition (πΉ 1 -score of 0.74 Β± 0.05
and 0.75 Β± 0.08 SR and BR, respectively), BR has fewer classes that
need classification (5 against 9 for bedroom and study, respectively),
meaning the actual performance of BR is lower. Thus, these results
imply that the bedroom is better for estimating the shower pres-
sures, likely due to less noise from the water and pipes and thus the
better signal propagation conditions, but that SR is better activity
recognition, likely do to the path crossing the actual activities.
The worst performing receiver is FB, with an πΉ 1 -score of 0.25 Β±
0.15, 0.90 Β± 0.01, and 0.47 Β± 0.28 for activity recognition, on/off,
and individual water pressure estimation, respectively. The worse
(a) results could be explained by additional noise caused from the fuse
(b) box itself (more of interference from pipes running in and out), or
that the most dominant signal is unaffected by the shower: from
Figure 3: Classification of water pressure based on 10 runs the transmitter to FB, it only needs to penetrate one door before it
with 250 epochs for π = {0, 1}, π = ππ
for the washing reaches the receiver, as the fuse box compartment is open at the
and idling activities (I,WHF-E/S,WBL-E/S), where (a) is binary top.
on/off and (b) identifies different water pressures.
6 CONCLUSION AND FUTURE WORK
5.1.2 Water pressure estimation. Figure 3 shows the confusion The results indicate that it is possible to determine whether the
matrices for the classification of the four water pressures for π = shower is on or off with an πΉ 1 -score of 0.99 Β± 0.00 and individ-
ππ
, π = 0. Figure 3 includes the actual showering activities (such as ual shower pressures with an πΉ 1 -score between 0.48 Β± 0.17 and
idling and washing), namely π π» πΉ βπΈ,π π» πΉ βπ,π π΅πΏ βπΈ,π π΅πΏ βπ. 0.79 Β± 0.24 based over 10 runs. This implies that while it is challeng-
Figures 3a,b are also normalized over the 10 runs. ing to detect certain individual shower pressures, it is feasible to
For the binary problem of estimating whether the shower is on monitor shower usage to a coarser degree. For activity recognition,
or off (Figure 3a), it is observable that this can be estimated with an πΉ 1 -score of 0.74Β±0.05 on average is found. However, it is possible
an πΉ 1 -score of 0.99 Β± 0 for the relevant activities (b). to differentiate between different levels (such as regular and slowed
For the individual shower pressures, the πΉ 1 -score of Off is 0.97 Β± down movements) of shower-related activities (such as washing
0.02 for classification with only the relevant activities. This is dif- head or body) with an πΉ 1 -score of 0.76 Β± 0.17 and 0.93 Β± 0.07. This
ferent for the other three: Low and Max have an πΉ 1 -score in the indicates channel state information can be used to potentially mon-
range of 0.73 Β± 0.19 β 0.79 Β± 0.24 for relevant activities, which is itor the personal hygiene for patients in self-care. Additionally, the
lower than for Off. This indicates that it is harder to differentiate results imply the optimal position to place the receiving receiver
between either. However, the lower πΉ 1 -score is largely explained for the activity recognition is directly behind the shower, while for
by investigating Med, with a πΉ 1 -score of 0.48 Β± 0.17: while Low and water pressure estimation it is recommended to put the receiver
Max have false-positives and false-negatives between them, this slightly skewed behind the shower for a more optimal propagation
is a minority compared to the false-positives and false-negatives environment.
CHIIoT 1, February 17, 2021, Delft, The Netherlands Klein Brinke, et al.
(a) (b) (c) (d)
Figure 4: Classification of activities and water pressure (10 runs, 250 epochs) for π = {0, 1} for receivers FB (a,c) and BR (b,d).
(a) and (b) show the activity recognition for relevant activities, while (c) and (d) shows the water pressure estimation.
Future work would include validating the results in this paper [10] Jeroen Klein Brinke and Nirvana Meratnia. 2019. Scaling Activity Recognition
through more participants, different locations for the transmitter, Using Channel State Information Through Convolutional Neural Networks and
Transfer Learning. 56β62. https://doi.org/10.1145/3363347.3363362
and different frequencies. Repeating the experiments with more [11] Melissa Krauss, Eileen Hitcho, Ngugi Kinyungu, William Dunagan, Irene Fischer,
(and different) participants for different settings (e.g. multi-floor Stanley Birge, Shirley Johnson, Eileen Costantinou, and Victoria Fraser. 2005.
A Case-Control Study of Patient, Medication, and Care-Related Risk Factors
houses or larger apartments) is a must to verify these results in on for Inpatient Falls. Journal of general internal medicine 20 (03 2005), 116β22.
a larger scale and for these technologies to be adapted into real-life https://doi.org/10.1111/j.1525-1497.2005.40171.x
scenarios. While 5 GHz has proven useful in this experiment, 2.4 [12] Hoonyong Lee, Changbum R. Ahn, and Nakjung Choi. 2020. Fine-grained oc-
cupant activity monitoring with Wi-Fi channel state information: Practical im-
GHz could be viable due to its larger wavelength, which means it is plementation of multiple receiver settings. Advanced Engineering Informatics 46
better capable at penetrating walls. Additionally, either frequency (2020), 101147. https://doi.org/10.1016/j.aei.2020.101147
could also be affected by water in a different way. The proposed [13] J. Liu, Y. Chen, Y. Wang, X. Chen, J. Cheng, and J. Yang. 2018. Monitoring Vital
Signs and Postures During Sleep Using WiFi Signals. IEEE Internet of Things
implementation should also be tested in a real-time for fashion for Journal 5, 3 (2018), 2071β2084. https://doi.org/10.1109/JIOT.2018.2822818
real-time classification. [14] Sameera Palipana, David Rojas, Piyush Agrawal, and Dirk Pesch. 2018. FallDeFi:
Ubiquitous Fall Detection Using Commodity Wi-Fi Devices. 1, 4 (2018). https:
//doi.org/10.1145/3161183
[15] Ferris S.H. de Leon M.J. Crook T. Reisberg, B. 1982. The Global Deterioration
REFERENCES Scale for assessment of primary degenerative dementia. The American journal of
psychiatry (9 1982), 1136β9. https://doi.org/10.1176/ajp.139.9.1136
[1] Romola Bucks, D.L. Ashworth, G.K. Wilcock, and K Siegfried. 1996. Assessment [16] S. A. Shah and F. Fioranelli. 2019. RF Sensing Technologies for Assisted Daily
of Activities of Daily Living in Dementia: Development of the Bristol Activities Living in Healthcare: A Comprehensive Review. IEEE Aerospace and Electronic
of Daily Living Scale. Age and ageing 25 (04 1996), 113β20. https://doi.org/10. Systems Magazine 34, 11 (2019), 26β44. https://doi.org/10.1109/MAES.2019.
1093/ageing/25.2.113 2933971
[2] Mhd Daher, Ahmad Diab, Maan El Badaoui El Najjar, Mohamad Khalil, and [17] Georgios Siantikos, Theodoros Giannakopoulos, and Stasinos Konstantopoulos.
François Charpillet. 2016. Elder Tracking and Fall Detection System Using Smart 2016. A Low-cost Approach for Detecting Activities of Daily Living using Audio
Tiles. IEEE Sensors Journal PP (11 2016), 1β1. https://doi.org/10.1109/JSEN.2016. Information: A Use Case on Bathroom Activity Monitoring. 26β32. https://doi.
2625099 org/10.5220/0005803700260032
[3] Giovanni Diraco, Alessandro Leone, and P. Siciliano. 2017. A Radar-Based Smart [18] Reto Stucki, Prabitha Urwyler, Luca Rampa, RenΓ© MΓΌri, Urs Mosimann, and
Sensor for Unobtrusive Elderly Monitoring in Ambient Assisted Living Applica- Tobias Nef. 2014. A Web-Based Non-Intrusive Ambient System to Measure and
tions. Biosensors 7 (11 2017), 55. https://doi.org/10.3390/bios7040055 Classify Activities of Daily Living. Journal of medical Internet research 16 (07
[4] G. Feng, J. Mai, Z. Ban, X. Guo, and G. Wang. 2016. Floor Pressure Imaging for 2014), e175. https://doi.org/10.2196/jmir.3465
Fall Detection with Fiber-Optic Sensors. IEEE Pervasive Computing 15, 2 (2016), [19] Wei Wang, Alex Liu, Muhammad Shahzad, Kang Ling, and Sanglu Lu. 2017.
40β47. https://doi.org/10.1109/MPRV.2016.27 Device-Free Human Activity Recognition Using Commercial WiFi Devices. IEEE
[5] Clarissa Giebel, Caroline Sutcliffe, and David Challis. 2014. Activities of Daily Journal on Selected Areas in Communications PP (03 2017), 1β1. https://doi.org/
Living and quality of life across different stages of dementia: a UK study. Aging 10.1109/JSAC.2017.2679658
mental health 19 (05 2014), 1β9. https://doi.org/10.1080/13607863.2014.915920 [20] Yan Wang, Jian Liu, Yingying Chen, Marco Gruteser, Jie Yang, and Hongbo Liu.
[6] C.M. Giebel, C. Sutcliffe, M. Stolt, S. Karlsson, A. Renom-Guiteras, M. Soto, 2014. E-eyes: Device-free location-oriented activity identification using fine-
H. Verbeek, A. Zabalegui, the RightTimePlaceCare Consortium, and D. Chal- grained WiFi signatures. Proceedings of the Annual International Conference on
lis. 2014. Deterioration of basic activities of daily living and their impact Mobile Computing and Networking, MOBICOM (09 2014). https://doi.org/10.1145/
on quality of life across different cognitive stages of dementia: a European 2639108.2639143
study. International Psychogeriatrics 26, 8 (1 Jan. 2014), 1283β1293. https: [21] Zhicheng Yang, Parth H. Pathak, Yunze Zeng, Xixi Liran, and Prasant Mohapatra.
//doi.org/10.1017/S1041610214000775 2017. Vital Sign and Sleep Monitoring Using Millimeter Wave. 13, 2 (2017).
[7] Daniel Halperin, Wenjun Hu, Anmol Sheth, and David Wetherall. 2011. Tool https://doi.org/10.1145/3051124
Release: Gathering 802.11n Traces with Channel State Information. Computer [22] Yue Zhang, Xiang Li, Shengjie Li, Jie Xiong, and Daqing Zhang. 2018. A Training-
Communication Review 41 (01 2011), 53. https://doi.org/10.1145/1925861.1925870 Free Contactless Human Vitality Monitoring Platform Using Commodity Wi-Fi
[8] S. Katz. 1983. Assessing self-maintenance: activities of daily living, mobility, and Devices. 488β491. https://doi.org/10.1145/3267305.3267613
instrumental activities of daily living. Journal of the American Geriatrics Society [23] Zizheng Zhang, Shigemi Ishida, Shigeaki Tagashira, and Akira Fukuda. 2019.
(12 1983), 721β7. https://doi.org/10.1111/j.1532-5415.1983.tb03391.x Danger-Pose Detection System Using Commodity Wi-Fi for Bathroom Monitor-
[9] S. S. Khan, T. Zhu, B. Ye, A. Mihailidis, A. Iaboni, K. Newman, A. H. Wang, and ing. Sensors 19 (02 2019). https://doi.org/10.3390/s19040884
L. S. Martin. 2017. DAAD: A Framework for Detecting Agitation and Aggression [24] D. Zhu, N. Pang, G. Li, and S. Liu. 2017. NotiFi: A ubiquitous WiFi-based abnor-
in People Living with Dementia Using a Novel Multi-modal Sensor Network. In mal activity detection system. In 2017 International Joint Conference on Neural
2017 IEEE International Conference on Data Mining Workshops (ICDMW). 703β710. Networks (IJCNN). 1766β1773. https://doi.org/10.1109/IJCNN.2017.7966064
https://doi.org/10.1109/ICDMW.2017.98