![The architecture of the proposed LSTM model (adopted from [23])](https://www.researchgate.net/profile/Raiyan-Jahangir/publication/379087334/figure/fig3/AS:11431281230586291@1711009139181/The-architecture-of-the-proposed-LSTM-model-adopted-from-23_Q320.jpg)
Content uploaded by Raiyan Jahangir
Author content
All content in this area was uploaded by Raiyan Jahangir on Mar 21, 2024
Content may be subject to copyright.
2023 IEEE 11th Region 10 Humanitarian Technology Conference
16-18 October, Rajkot, Gujarat, India
Development of a Smart Infant Monitoring System
for Working Mothers
Raiyan Jahangir1, Nasif Shahriar Mohim, Afnan Alauddin Mumu, Mahapara Naim,
Anika Ashraf, Md Abu Syed, Muhammad Nazrul Islam2
Department of Computer Science and Engineering, Military Institute of Science and Technology,
Mirpur Cantonment, Dhaka-1216, Bangladesh
Email: 1raiyan.jahangir@gmail.com, 2nazrul@cse.mist.ac.bd
Abstract—Infants must be monitored around the clock because
they are at risk for various health issues (breathing difficulties
and feeding problems) and cannot communicate their needs
properly. This becomes more essential for mothers working
outside, keeping their child or infants at home or a daycare
center. Though several infant monitoring systems exist to mon-
itor infants, most of these do not have all possible intelligent
features like detecting infant cries, which is the primary form of
infant communication. Therefore, this research offers a machine
learning-based smart infant monitoring system for working
mothers. The proposed system can provide the infant’s pulse
rate, temperature, location, and alerts for crying and abnormal
health conditions. The system was also evaluated with 15 working
mothers and found that the proposed system could provide an
easier and more convenient way to monitor their infants.
Keywords—Cry detection, machine learning, infant monitor-
ing, usability, usability evaluation, and Long Short-Term Mem-
ory.
I. INTRODUCTION
An infant is a child aged between birth and a year old.
Infants grow and develop at a fantastic rate from birth to
one year of age. They pick up skills like smiling, waving,
clapping, picking up items, rolling over, sitting up, crawling,
and babbling; some even start pronouncing a few words.
They grow to trust and form bonds with their caregivers and
frequently comprehend more than they can express [1].
Infants’ vulnerability to diseases requires frequent moni-
toring of their pulse rate, heartbeat, body temperature, and
breathing [2]. However, constantly monitoring them is chal-
lenging, especially for working mothers. Hiring caregivers or
using daycare facilities is not always convenient or reliable. To
alleviate stress and worry, an infant monitoring system enables
working mothers to monitor their infants, providing a practical
solution remotely.
An infant monitoring system is like an alarm system
that can track infants’ movements, activities, and different
health parameters and alert the appropriate authorities through
alarms, radio, mobile phones, or any display. Since the dawn
of humanity, families have had the instinct to protect their
infants from hazards and dangers [3]. However, how working
mothers raise their infants has changed due to technological
advancements [4]. As such, a smart infant monitoring system
might be a solution for them.
Most of the infant monitoring systems developed have
mainly focused on monitoring the health parameters of infants.
Most of the time, an infant usually tries to get the attention of
their mothers or caregivers by crying; thus, detecting when an
infant cries is also necessary. Some previous work on infant
monitoring systems used a sound sensor to detect infant cries
[5]. But this type of system takes action whenever any high-
intensity sound occurs. Some of the studies used a camera
with face recognition and image processing [6] features along
with sound sensors [7]. These systems simultaneously check
the facial expressions of infants and sense sound. Whenever
any sound is heard, the systems check the facial expressions of
the infants. If the facial expression is like a crying infant, only
then does the system provide an alert to the mother. These
systems work with video streams that take more bandwidth
and require more costly setups. Detecting infant cries through
machine learning could be an effective solution. Again, very
few of the previous works had any functionalities to track
the location of infants. It might be necessary to keep their
locations tracked to ensure their safety. Previous research on
infant monitoring systems has not always included proper
evaluation of the system’s accuracy and usability [8], [9].
This is problematic, as it can lead to unreliable results and
interventions. To address this issue, researchers are turning to
ensure accuracy and usability [10] for developing systems for
monitoring infants.
Therefore, this research aims to propose a conceptual
framework to develop an infant monitoring system to assist
working mothers. Secondly, to develop an infant monitoring
system incorporating machine learning, sensors, and a mobile
application to solve the problems the previous works did not
address. Finally, to evaluate the performance of the developed
system.
The remaining portions of the paper are organized as fol-
lows. The related works are discussed in Section II. Section III
covers the system development. The evaluation procedure, data
analysis, and findings are discussed in Section IV. Concluding
remarks with the research limitation and scopes for future
work are presented in Section V.
II. LITERATURE REVIEW
Various solutions are available for computers, smartphones,
and tablets which assist working mothers in keeping an eye
on their children. The first concept of an infant monitoring
system was given by Singh and Hsiao [11], which describes
979-8-3503-2614-7/23/$31.00 ©2023 IEEE
how the monitoring system can communicate with working
mothers remotely.
A few studies based on Arduino microcontroller was de-
veloped to monitor infants in intensive care units and protect
them from sudden death. Palaskar et al. [12] presented a novel
method of an automatic monitoring system for infant care. The
authors created a low-cost infant monitoring system to hear
an infant’s cry. In this system, the cradle automatically swings
when the system senses sound, continuing to swing until the
infant stops crying. Additionally, a camera is fixed to the top
of the cradle to record footage of the infant’s surroundings.
Ishak et al. [18] developed a system with sensors attached to
an incubator to gauge heart rate and humidity. An Arduino
device interface transmits this information to a computer for
storage. The system of Gupta et al. [15] uses sensors that
may be attached to the infant’s clothing. Electrocardiography
(ECG), temperature, and carbon dioxide (CO2) sensors are
integral to this system and are placed around the infant’s bed.
This health monitoring device can be utilized at home and in
Neonatal Intensive Care Units (NICU), providing a real-time
signal of infant status changes. Patil et al. [14] designed a
GSM-based smart infant monitoring system. Important factors,
including body temperature, humidity level, pulse rate, and
mobility of an infant, are monitored by this system, which
uses a GSM network to transmit the data to their mothers.
The working mothers can be alerted with details to perform
any necessary action. The sensors for monitoring significant
parameters, LCD, GSM interface, and sound alert are all
controlled by a single microcontroller in the system design.
Eventually, infant monitoring systems were implemented
with Raspberry Pi, which gives better flexibility and more
functionalities in a monitoring system. The system of Patel
et al. [16] uses a Raspberry Pi to connect a static camera,
motors, and other infrared sensors. The camera is positioned
to capture the whole room. Working mothers can log into a
website to access any information or updates. The system
can also stream videos of the infants online. An Arduino
board is also connected to a GSM module that sends alerts
anytime the camera rotates. Annouri et al. [13] made a stand-
alone integrated monitor placed on top of the infant’s bed,
incorporating several functions offered by separate wireless
monitors. There are sensors for temperature, motion, and
sound to prevent unintentional choking and suffocation of
infants in bed. The infant’s temperature, audio, and video are
all accessible by a smartphone. In the system of Pai et al.
[17], working mothers can see and monitor their infants from
a distance and check parameters like humidity, temperature,
and movements of infants. An infant’s sleep and sleep patterns
are also automatically recorded. A camera is employed for the
observer and the working mothers to see it. The system sends
the parameter’s specifics to the working mothers so they can
respond if anything happens. Symon et al. [19] proposed a
system that consists of a condenser microphone to record the
infant’s cries, a PIR motion sensor to catch the infant moving,
and a Pi camera to record the infant’s motion. The equipment
is connected to a display to output the baby’s sound and video.
This review of previous works shows that each system
supports some functionalities while missing other important
ones. Some of the systems have the functionality of detecting
infant cries. The systems either used video streaming to see
the infants’ facial expressions or sound sensors to detect
sound. But, in the former, the facial recognizer in videos may
wrongly consider an infant’s expression to be crying. In the
case of sound sensors, any sound might be perceived as an
infant cry. Again, a few of the systems have functionalities
for tracking the location of infants. For security purposes, it
is necessary to continuously track the infant’s location and
deliver it to the working mothers. This research aims to
overcome these problems in the existing systems and make
a better infant monitoring system. This system will work with
audio data to detect infant cries with neural networks and alert
working mothers only when an infant cries. It will also have
functionalities to track the infants’ location, sense their health
parameters, and notify their mothers when necessary. However,
a comparison of the previous systems to the proposed system
is given in table I.
III. SYS TE M DEV ELOPMENT
A. Conceptual Framework
The conceptual framework of the proposed system is illus-
trated in figure 1. A Raspberry Pi is used as the processor
of the proposed system to connect other system devices. The
system was divided into three modules: Data Acquisition
Module, Application Module, and Cry Detection Module.
The Data Acquisition Module consists of different sensors
like the temperature sensor, pulse rate sensor, GPS unit, and
microphone to collect data from infants in real-time, process
them, and then store them in the database. The Application
Module consists of a mobile application that the working
mothers will use to receive the data from the database. The Cry
Detection Module detects whether an infant cries whenever a
sound is heard. The module was deployed in the Raspberry
Pi. For storing the infant data, the Firebase cloud database is
used.
Figure 2 shows the workflow diagram of the proposed
system. The device must be powered on using a power bank
or battery. The LM35 sensor measures body temperature, the
pulse sensor measures the pulse rate, and the GPS unit locates
the infant. The sensors send their data to the Raspberry Pi.
The data is then sent to Firebase. According to a study [20],
the normal range of pulse rate for an infant of 0-3 months is in
the field of (100-150) BPM, and for infants of 3-12 months,
it is (80-120) BPM. The usual body temperature range for
infants is (36.4-37.9)°C [21]. If the data of either pulse rate
or temperature or both are found to be more or less than the
normal range of an infant, it will be considered as “Abnormal
Data” by the system. When temperature or pulse rate is found
to be abnormal, Firebase notifies the mothers. Again, when
the infant cries, the crying sound will be recorded by the
microphone and detected using a neural network model. If the
model identifies the sound as an infant cry, a notification is sent
to the working mothers that the infant is crying. The infant’s
TABLE I: Comparison between existing and proposed system
System developed by Can sense
health parameters
Can track
location
Detects
sound
Uses a camera to
detect infant cry
Uses infant cry detection
with Machine Learning
Alerts parents for health
and cry alerts of infants
Uses mobile
application
Palaskar et al. [12] No No Yes Yes No No No
Annouri et al. [13] Yes No Yes Yes No Yes Yes
Patil et al. [14] Yes No Yes No No Yes No
Gupta et al. [15] Yes No No No No Yes Yes
Patel et al. [16] No No No Yes No Yes No
Pai et al. [17] Yes No Yes Yes No Yes No
Ishak et al. [18] Yes No No No No Yes Yes
Symon et al. [19] Yes No Yes Yes No No No
Proposed system Yes Yes Yes No Yes Yes Yes
Fig. 1: System architecture of the proposed system
location, temperature, and pulse rate data stored in Firebase
are then displayed in the mobile app. The flow continues till
the device is running.
In the following subsections, each module is presented in
more detail:
B. Data Acquisition Module
The pulse rate sensor senses the pulse rate of an infant in
Beats Per Minute (BPM). The LM35 sensor senses the infant’s
body temperature per Degree Celcius (°C). The GPS unit
connects itself to nearby satellites and then gives the latitude
and longitude of its current location as output. The mini
microphone records infant cries. All these acquired data are
then sent to the Raspberry Pi for further processing. Finally,
Raspberry Pi sends these data to the database.
C. Application Module
A simple yet useful mobile application was developed
using Flutter. The target users of this application are working
mothers who want to monitor their infants. The users get
various functionalities on the mobile application, as seen in
figure 3a. Some of them are described below:
Fig. 2: Workflow diagram of the proposed system
1) Displaying Health Graph:This page displays the in-
fant’s pulse rate and body temperature in the form of a line
graph (illustrated in figure 3b). There is another page that
shows the data in text form. This functionality helps working
mothers get a visual understanding of their infants’ current
health.
2) Displaying Location on Map:This page displays the
current location of the infants (illustrated in figure 3c). This
functionality can let working mothers know about their infants’
whereabouts. The working mothers can immediately enquire
if their infants’ location is shown elsewhere.
3) Displaying Notifications:This page shows all the noti-
fications or alerts in a list. “Health Alert” comes whenever an
infant’s pulse rate or body temperature is more or less than the
normal range. “Cry Alert” comes when the system detects an
infant crying. The notifications received are listed along with
the time and reason (illustrated in figure 3d).
D. Cry Detection Module
The cry detection module consists of a deep recurrent
neural network model. Neural networks are a type of machine
learning algorithm that works like a human brain [22]. Long
Short-Term Memory (LSTM) neural networks have been used
(a) Home page (b) Page showing health graph (c) Page showing location (d) Page showing notifications
Fig. 3: Various pages of the mobile application
for the model training in this study. This approach was chosen
because it was known from previous literature that it is
appropriate for audio classification [23].
1) Data Collection:The dataset created by Piczak [24] was
used in this study. A total of 432 audio recordings are in the
dataset. The recordings include a variety of sounds, including
crying, laughing, noise, and silence.
2) Data Preprocessing and Feature Extraction:The audio
recordings were initially loaded, and sampled at 44100 Hz,
with each given a duration of 5 seconds. If the audio is longer
than 5 seconds, it is split into additional samples with a 5-
second duration. The audio file is padded to maintain the same
duration if the duration is less than 5 seconds. Each audio
file was then transformed into a mel-scaled spectrogram. Each
spectrogram yielded 25 Mel-Frequency Cepstrum Coefficients
(MFCCs) [25]. The average MFCC values across the audio
data were then calculated using the means of the coefficients,
and a single feature vector was produced.
3) Developing the model:An RNN variation, LSTM, intro-
duces a novel structure known as a memory cell. It comprises
four essential components: an input gate, a neuron with a self-
recurrent connection (a link to itself), a forget gate, and an
output gate. The LSTM neuron differs from a conventional
one because it has three new gates. The primary idea is to use
these gates to prevent the gradient from diverging; they can
choose what to store, read, and write and when to do so.
The proposed architecture, shown in figure 4, comprises
an input layer that takes in pre-processed signals as input,
two recurrent layers with LSTM cells plus a dense layer, a
fully connected layer where the features from the previous
layer are given as input, and a softmax activation function
for the final classification. The whole network is trained by
minimizing the categorical cross-entropy as a loss function.
The network is trained with four classes: “cry”, “laugh”,
“noise”, and “silence”.
Fig. 4: The architecture of the proposed LSTM model (adopted
from [23])
TABLE II: Performance measures for the developed models
Accuracy
(train)
Precision
(train)
Recall
(train)
F1 score
(train)
Accuracy
(test)
Precision
(test)
Recall
(test)
F1 score
(test)
87.3 87.3 87.5 87.3 91.3 91.8 88.5 89.3
4) Evaluating the model:In this study, the model was
generated using the train dataset, whereas the model’s perfor-
mance is evaluated for both the train and the unknown dataset
(test set). The model’s performance was measured in terms of
accuracy, precision, recall, and f1-score, and the results are
shown in Table II. Table II shows that the accuracy, precision,
recall, and f1-score for both the train and test data are more
than 87%, which is ideal for a model.
Figure 5 shows the final integrated circuit and the final
prototype in the form of an infant belt. The circuit is put inside
the infant belt. The belt will be worn by the infant that will
be monitored.
Fig. 5: Final Prototype of the proposed system
IV. SYS TE M EVALUATION
The system was evaluated to assess its usability following
the methodology adopted in [26]–[30].
A. Participant’s Profile
15 participants who are working mothers and have infants
aged six months to 1 year participated in this case study.
The participants’ average age was approximately 33 years and
ranged between 28 to 38 years. They all have mobile phones
and know how to use mobile apps. They have consented to let
their infants wear the belt containing the proposed monitoring
system. However, none of them had any previous idea about
the system, nor did they participate in the stages of requirement
elicitation and system development.
B. Study Procedure
The participants were briefed about the purpose of this user
testing as part of usability evaluation, and they were informed
that it is volunteer participation. Written consent was also
taken from them in this user study, and their biographical data
were taken. Then a demonstration of the system was shown
and briefed to the participants.
The participants were initially provided instructions regard-
ing the overall system and their working procedure. They were
given the smartphone application to test the functionalities, and
the infants wore the belt containing the system. Then they
were given to do a set of tasks, mentioned in table IV. Before
formally collecting the data, they were allowed to explore
the system functionalities for 15-20 minutes. The system’s
effectiveness was determined by measuring the accuracy of
the hardware functionalities. The evaluation was divided into
5 functionalities, with 10 attempts on each functionality.
The system’s efficiency was determined by letting the
participants use the mobile application. The participants were
observed while they were carrying out the required tasks.
Finally, after completing the user study, questionnaires were
provided to the participants to get their overall feedback about
TABLE III: Accuracy of different functionalities of the system
Feature No. and Description Average Success rate Delay(M±SD) (s)
F1: Pulse Rate Sensing 100% 0±0.0
F2: Temperature Sensing 100% 0±0.0
F3: Location Tracking 80% 0±0.0
F4: Cry Detection and Alert 90% 1±0.75
F5: Health Alert 100% 1±0.69
the developed system (Table V). Both objective and subjec-
tive data were collected to assess the system’s usability and
performance with effectiveness, efficiency, and satisfaction.
C. Result Analysis
The summary results are briefly presented in the tables III-
V. It can be observed from the table III that functionality
F1, F2, and F5 works perfectly (100%) without any errors.
Functionality F3 and F4 work almost perfectly (80% and 90%,
respectively). Thus, the effectiveness of our system is up to
the mark.
It is observed from table IV that tasks T1, T2, T4, and T5
were completed successfully by each participant in one attempt
with no request for help. However, 2 participants asked for
help while conducting task T2. Task T3 took longer than all
the other tasks. 3 participants attempted this task twice and
asked for help to complete the task. It is assumed that the
different navigation routes of task T3 compared to tasks T1
and T2 resulted in this confusion. Other than that, they could
complete all the tasks with ease.
Table V shows the average ratings of each question on a
scale of 5. It is seen that most of the participants expressed
satisfaction with using the system. They enjoyed using the
system, and all agreed to recommend the developed system to
others.
Thus, it can be concluded that the proposed system has been
highly accurate, functional, and successful.
V. CONCLUSIONS
The proposed system uses sensors to detect health parame-
ters and track an infant’s location for monitoring. A neural net-
work model is also adopted in the system, which was trained to
detect the cries of infants and notify the infants’ mothers and
other authorized guardians via mobile notifications. An infant
belt was used as a wearable device, while the working mothers
used the mobile application to monitor their infants. There are
some limitations of the proposed system. Firstly, the proposed
system requires internet connectivity to send monitoring data
to a remote server. Secondly, the proposed system can detect
an infant’s cry and alert working mothers but cannot determine
why the infant is crying. Thirdly, the dataset used to train the
neural network models is not very large. Finally, the number
of participants involved in the user study was not adequate.
Future work will focus on these limitations to develop a more
effective and usable infant monitoring application.
TABLE IV: Measures of different parameters for usability evaluation
Task No. and Description Average Task Completion
Time (seconds)
No. of
Optimal Clicks
Average no. of
clicks taken
Average no.
of attempts
Average no. of times
asking for help
Average
Success Rate
T1: Logging in to the system 3.2 1 1.2 1 0 100%
T2: Registering an infant. 5.4 2 2.2 1 0.13 100%
T3: Checking infant Location 7.6 1 1.4 1.2 0.2 90%
T4: Checking Notifications 3 1 1 1 0 100%
T5: Checking Health Parameters of infants 3.8 2 2.6 1 0 100%
TABLE V: User feedback to measure satisfaction
Data Type Average Ratings Mean and SD
Overall Satisfaction 5/5 5±0.0
Easy to Use 4.8/5 4.8±0.18
Easy to Learn 4.67/5 4.67±0.27
Future Use 5/5 5±0.0
Recommend Others 5/5 5±0.0
REFERENCES
[1] K. E. Adolph and S. E. Berger, “Physical and motor development,” in
Developmental Science. Psychology Press, 2015, pp. 269–342.
[2] P. McDougall and M. Harrison, “Fever and feverish illness in children
under five years.” Nursing Standard, vol. 28, no. 30, 2014.
[3] D. E. Eyer, “Mother-infant bonding: A scientific fiction,” Human Nature,
vol. 5, no. 1, pp. 69–94, 1994.
[4] S. Brangui, M. El Kihal, and Y. Salih-Alj, “An enhanced noise cancelling
system for a comprehensive monitoring and control of baby environ-
ments,” in 2015 International conference on electrical and information
technologies (ICEIT). IEEE, 2015, pp. 404–409.
[5] Y. K. Dubey and S. Damke, “Baby monitoring system using image
processing and iot,” International Journal of Engineering and Advanced
Technology, vol. 8, no. 6, pp. 4961–4964, 2019.
[6] R. Jahangir, M. Wasif-Ul-Islam, N. S. Mohim, A. Ashraf, N. I. Khan,
and M. N. Islam, “Towards developing a voice-over-guided system
for visually impaired people to learn writing the alphabets,” in 2022
25th International Conference on Computer and Information Technology
(ICCIT). IEEE, 2022, pp. 1015–1020.
[7] D. M. Ibrahim, M. A. A. Hammoudeh, S. Ambreen, and S. Mohammadi,
“Raspberry pi-based smart infant monitoring system,” International
journal of engineering research and technology, vol. 12, no. 10, pp.
1723–1729, 2019.
[8] M. A. Razzak, M. N. Islam, T. Broti, E. S. Kamal, and S. Zahan,
“Exploring usability problems of mhealth applications developed for
cervical cancer: An empirical study,” SAGE Open Medicine, vol. 11, p.
20503121231180413, 2023.
[9] M. N. Islam, M. M. Karim, T. T. Inan, and A. Islam, “Investigating
usability of mobile health applications in bangladesh,” BMC medical
informatics and decision making, vol. 20, no. 1, pp. 1–13, 2020.
[10] N. Bevan, J. Carter, and S. Harker, “Iso 9241-11 revised: What have we
learnt about usability since 1998?” in Human-Computer Interaction: De-
sign and Evaluation: 17th International Conference, HCI International
2015, Los Angeles, CA, USA, August 2-7, 2015, Proceedings, Part I 17.
Springer, 2015, pp. 143–151.
[11] S. Singh and H. Hsiao, “Internet based infant monitoring system,” in
Proceedings of the First Joint BMES/EMBS Conference. 1999 IEEE
Engineering in Medicine and Biology 21st Annual Conference and the
1999 Annual Fall Meeting of the Biomedical Engineering Society (Cat.
N, vol. 2. IEEE, 1999, pp. 674–vol.
[12] R. Palaskar, S. Pandey, A. Telang, A. Wagh, and R. M. Kagalkar, “An
automatic monitoring and swing the baby cradle for infant care,” Interna-
tional Journal of Advanced Research in Computer and Communication
Engineering, vol. 4, no. 12, pp. 187–189, 2015.
[13] J. Annouri, T. Moyo, H. Wang, D. Zuber, Z. Soriano, and D. Soriano,
“Knight’s wireless baby monitor,” 2014.
[14] S. P. Patil and M. R. Mhetre, “Intelligent baby monitoring system,” ITSI
Transactions on Electrical and Electronics Engineering, vol. 2, no. 1,
pp. 11–16, 2014.
[15] S. Gupta, Z. M. Khan, R. Srivastava, and P. Chougule, “Infant monitor-
ing system using multiple sensors,” International Journal of Research in
Engineering and Technology (IJRET), vol. 5, no. 5, pp. 337–339, 2016.
[16] P. B. Patel, V. M. Choksi, S. Jadhav, and M. Potdar, “Smart motion
detection system using raspberry pi,” International Journal of Applied
Information Systems (IJAIS), vol. 10, no. 5, pp. 37–40, 2016.
[17] S. S. Pai, S. C. Pereira, T. Nicole, and A. Ushadevi, “Advanced baby
monitor,” International Journal of Internet of Things, vol. 6, no. 2, pp.
51–55, 2017.
[18] D. N. F. M. Ishak, M. M. A. Jamil, and R. Ambar, “Arduino based
infant monitoring system,” in IOP conference series: materials science
and engineering, vol. 226, no. 1. IOP Publishing, 2017, p. 012095.
[19] A. F. Symon, N. Hassan, H. Rashid, I. U. Ahmed, and S. T. Reza,
“Design and development of a smart baby monitoring system based on
raspberry pi and pi camera,” in 2017 4th International Conference on
Advances in Electrical Engineering (ICAEE). IEEE, 2017, pp. 117–
122.
[20] F. O’Leary, A. Hayen, F. Lockie, and J. Peat, “Defining normal ranges
and centiles for heart and respiratory rates in infants and children: a
cross-sectional study of patients attending an australian tertiary hospital
paediatric emergency department,” Archives of disease in childhood, vol.
100, no. 8, pp. 733–737, 2015.
[21] J. l. Takayama, W. Teng, J. Uyemoto, T. B. Newman, and R. H. Pantell,
“Body temperature of newborns: what is normal?” Clinical Pediatrics,
vol. 39, no. 9, pp. 503–510, 2000.
[22] C. M. Bishop, “Neural networks and their applications,” Review of
scientific instruments, vol. 65, no. 6, pp. 1803–1832, 1994.
[23] M. Scarpiniti, D. Comminiello, A. Uncini, and Y.-C. Lee, “Deep re-
current neural networks for audio classification in construction sites,” in
2020 28th European Signal Processing Conference (EUSIPCO). IEEE,
2021, pp. 810–814.
[24] K. J. Piczak, “ESC: Dataset for Environmental Sound Classification,” in
Proceedings of the 23rd Annual ACM Conference on Multimedia. ACM
Press, pp. 1015–1018.
[25] B. Logan et al., “Mel frequency cepstral coefficients for music model-
ing.” in Ismir, vol. 270, no. 1. Plymouth, MA, 2000, p. 11.
[26] M. N. Islam, M. A. Ahmed, and A. N. Islam, “Chakuri-bazaar: A mobile
application for illiterate and semi-literate people for searching employ-
ment,” International Journal of Mobile Human Computer Interaction
(IJMHCI), vol. 12, no. 2, pp. 22–39, 2020.
[27] N. Hasan, M. N. Islam, and N. Choudhury, “Evaluation of an interactive
computer-enabled tabletop learning tool for children with special needs,”
Journal of Educational Computing Research, vol. 60, no. 8, pp. 2105–
2137, 2023.
[28] N. I. Khan, S. N. Mustafina, F. F. Jhumu, A. Zobyer, M. H. Mahin,
M. A. I. Tarek, R. Rahman, and M. N. Islam, “Towards developing an
automated attendance management system using fingerprint sensor,” in
2020 Emerging Technology in Computing, Communication and Elec-
tronics (ETCCE). IEEE, 2020, pp. 1–6.
[29] M. A. Ahmed, M. N. Islam, F. Jannat, and Z. Sultana, “Towards
developing a mobile application for illiterate people to reduce digital
divide,” in 2019 International Conference on Computer Communication
and Informatics (ICCCI). IEEE, 2019, pp. 1–5.
[30] M. N. Islam, U. Hasan, F. Islam, S. T. Anuva, T. Zaki, and A. N.
Islam, “Iot-based serious gaming platform for improving cognitive skills
of children with special needs,” Journal of Educational Computing
Research, vol. 60, no. 6, pp. 1588–1611, 2022.
