Vous êtes sur la page 1sur 5

Smart Monitor for Tracking Child’s Brain Development

T. Anparasanesan K. Mathangi S. Kobikanth


Faculty of Computing Faculty of Computing Faculty of Computing
Sri Lanka Institute of Information Sri Lanka Institute of Information Sri Lanka Institute of Information
Technology (SLIIT) Technology (SLIIT) Technology (SLIIT)
Malabe, Sri Lanka Malabe, Sri Lanka Malabe, Sri Lanka
+94 774962290 +94 776841387 +94 776447004
anparasanesan@gmail.co mathangi100@gmail.com skobikanth@gmail.com
m
S. Seyon M. P. A. W. Anjalie
Faculty of Computing Gamage
Sri Lanka Institute of Information Faculty of Computing
Technology (SLIIT) Sri Lanka Institute of Information
Malabe, Sri Lanka Technology (SLIIT)
+94 776487999 +94 776028400
seyon94@gmail.com anjalie.g@sliit.lk

ABSTRACT engine. The system generates a report and notifies parents about
This paper provides a way to track the brain development of their child’s statistics periodically. This paper elaborates the
children and improving it via gamification. Machine Learning and procedure of model development, model training, model testing
Gamification are the key technologies used here. As the and development of suitable brain games in details. The results of
population rises, the demand for cost-effective methods to reduce the research work and future works are also discussed in the
the rate of cognitive decline becomes higher. A mobile following sections.
application is developed to track and develop the brain
development of children. In the mobile application, the child CCS Concepts
initially undergoes an evaluation phase to determine the current Human-centered computing~Ubiquitous and mobile
level of the cognitive skills of the child. Milestones particular to computing design and evaluation methods
that age category are also tracked in this evaluation phase. The
results of this evaluation phase are analyzed by the machine Keywords
learning model and suitable brain games are suggested. K-means android application; children; clustering; cognitive skills;
algorithm is used to develop the model which is an unsupervised gamification; k-means algorithm; machine learning
learning algorithm. The dataset is prepared by storing the results
of each game category in the evaluation phase. Data
preprocessing is done to clean up the dataset. During this period,
1. INTRODUCTION
Neurological research shows that the early years play a key role
data undergoes a series of steps. The dataset is divided into 80%
in children’s brain development[1]. A child’s brain undergoes an
and 20%. 80% of the dataset is used as the training dataset and the
amazing period of development during the first six years.
remaining 20% as the test dataset. The accuracy of the model is
Children cultivate 85% of their intellect, personality, and skills by
checked several times against the test data. Model accuracy is
age five[2]. So, it is vital to pay enough attention in the early
improved through model training and finally, the model got an
years to track and develop brain development. But in developing
accuracy of 88.49%. For the child, proper training is given to
countries, more than 200 million children under five years fail to
improve his cognitive skills and thus the brain development using
reach their potential in cognitive and social development[3].
Gamification. Games are developed using the UNITY game
The world is becoming mechanized, and the people are running
Permission to make digital or hard copies of all or part of this work for
out of time. Though it is important to track the child’s brain
personal or classroom use is granted without fee provided that copies are
not made or distributed for profit or commercial advantage and that development and provide necessary guidance and training they
copies bear this notice and the full citation on the first page. Copyrights need, parents find it hard to tend to their children as parents do
for components of this work owned by others than ACM must be not have enough time as they are occupied and busy. This might
honored. Abstracting with credit is permitted. To copy otherwise, or lead the child into loneliness unless the child has grandparents or
republish, to post on servers or to redistribute to lists, requires prior a babysitter. Moreover, this loneliness can be a threat to the
specific permission and/or a fee. psychology of the child and can affect the mental balance and
Request permissions from Permissions@acm.org. thus the brain development. This might lead them totally on a
ICEIT 2019, March 2–4, 2019, Cambridge, United Kingdom
different path in their future which might make them
© 2019 Association for Computing Machinery.
ACM ISBN 978-1-4503-6267-2/19/03…$15.00 antisocial/introverts.
http://doi.org/10.1145/3318396.3318401 So, considering all these problems, an application is developed
which is capable of tracking a child’s brain development and

68
improves it via gamification. Machine Learning and particular reward, even if that reward is something as simple as
Gamification are the key technologies used to develop the moving on to the next level[6].
application. Machine learning is used successfully in many
intellectual applications. The developed system is an Android Why Gamification
application with multiple features. Parents can easily track their Since 2014, gamification - the use of game design elements in
children’s brain development with this application ‘MAKS’. non-game contexts – has seen rapid adaption in the software
industry as well as a growing topic for research on its uses and
Nowadays, the influence of technology on children and education effects.
has been immense. These days children are very comfortable
with electronic gadgets and equipment. To keep the children Successful Gamification tools work because they [7]:
interested, using a gamification based mobile application for a. Give users the motivation to do something.
education may assist. b. Give users the ability to carry out a task.
It is true that children are more interested in playing games rather c. Give the user a trigger to complete the action.
than studying. So, it is an advantageous idea to develop their Paul Howard-Jones at the University of Bristol has been
cognitive skills at a very prominent time via gamification. This conducting fMRI studies on gamification in learning for several
research work is aimed at children to develop their cognitive years. His findings demonstrate something of significance that
skills via gamification. The machine learning model used here colleagues at the University of Toronto's Faculty of Medicine are
suggests the suitable brain games and their levels to each child, just beginning to investigate, and they have shown and proven the
analyzing their base cognitive skills level. activation of the ventral striatum, the central reward and learning
In this research work, a mobile application is developed to track a center of the brain, to be triggered by game mechanics but not
child’s brain growth level and to improve it via gamification. The through self-study or quiz completion [8].
proper ways to test the brain growth factors like concentration, Therefore, it is planned to use gamification technique to give
memory power, response time, attention are discovered in the training to children to improve the skills in which they lag.
Evaluation phase. Evaluation phase is a testing environment for
children in the Android app to determine their basic level of
cognitive skills. Memory, Concentration, Attention and Response 2. RELATED WORK
time are some factors that occupy a prominent place in brain There are many apps available for children, but they are very
development. Those factors are tested in this phase for each child generic. They do not evaluate each child individually and suggest
through appropriate games and their scores are stored in Firebase. games according to their status. Further, those apps do not
The dataset used to cluster is built up from this, respectively four guarantee the improvement of a child’s cognitive skills. The
columns each contains scores of respective games. Additionally, available apps are very generic gaming applications developed for
milestones are also tracked against age in the Evaluation phase. entertainment purpose only. And most of the brain-boosting
Developmental milestones are behaviors or physical skills seen in applications are not free. Only a limited version is free. They cost
infants and children as they grow and develop. The milestones considerably a large amount. In this work, all these missing
are different for each age range[4]. The standards of the features are added and an app is developed which is more
milestones are collected from a recognized pediatrician. interesting and specific to children. This application also checks
out the milestones specific to age. The user can use this
The child undergoes an Evaluation phase initially to determine application for free once it is hosted in the play store. And this
the level of cognitive skills and to track the milestones. Then in mobile application is not similar to other generic gaming
the decision-making process, the system finds out where the child applications because in other applications, the game starts from
stands in between the relevant range of brain growth by analyzing level 1, but here level will be selected automatically according to
the collected data, i.e., their scores in each game. Then the the status of the child.
machine learning model suggests the most suitable games to play
to develop the skill in which the child lags. The system interacts
with the child and trains it accordingly to increase his brain
3. METHODOLOGY
growth level via gamification. The application also maintains Following explain how the system works.
statistics about each child. It generates a report periodically,  Initially, the system evaluates the child to determine the
stating the child’s brain growth status and suggestions and current level of the child’s cognitive skills and to track the
notifies parents, so parents can interact with the child in a way the milestones.
child needs. This is how the system works.
 The results of the evaluation phase are used to determine
K-means clustering is the algorithm used here to build the model. whether the child has attained the expected brain growth for
It is an unsupervised learning algorithm, which is used when the his/her age.
dataset is unlabeled (i.e., data without defined categories or  A dataset is prepared based on the results from the testing
group). The objective of this algorithm is to find groups in the environment containing the values for the four factors
data, with the number of groups represented by the variable K[5]. mentioned earlier.
Gamification has become increasingly popular since it offers a  A model is developed and well trained to suggest suitable
wide range of advantages and helps to make the overall brain games based on child’s data collected from the testing
experience not only more enjoyable but more effective too. When environment.
people participate in activities that stimulate their bodies or minds,  A child plays the suggested brain games for a certain period
such as exercising, our body releases a hormone known as of time as directed. And later when he undergoes the
endorphins. The same effect can be achieved by playing games Evaluation phase again and re-evaluates himself, the
that challenge learners or give them the chance to achieve a

69
increase in scores indicates the improvement in his cognitive place in brain growth. Concentration, response time, memory and
skills and thus the brain development. attention are the factors to be tested on children. So, to evaluate
children through these factors to determine their current level of
The proposed system is developed as an Android application. cognitive skills, an Evaluation phase is designed in the Android
The Android Operating System is selected because using a app. The milestones are also checked out against age in phase.
machine learning model along with a variety of games in a cross-
platform is complex. Android has reachability among the public, As children are the key users of the application, an interactive
cost-effective and comparatively easy to integrate with the environment (UI/UX) is discovered to keep them interested.
UNITY engine. UNITY has been used to design and develop
games in this project. 3.2 Developing a model for prediction
To create a model, a suitable algorithm has to be selected
according to the specifications. Clustering needs to be done here
to the dataset. So, initially, four different clustering algorithms
were selected and tested against the dataset to find out the model
with higher accuracy. The accuracy of the model depends on the
machine learning algorithm that is used to build up the model and
the training provided to the model. So, by analyzing all four
algorithms, one algorithm needs to be selected precisely. The
selected four clustering algorithms are:
I. Self-Organizing Map (SOM)
II. K-Means Clustering algorithm (K-Means)
III. Expectation Maximization Clustering (EM)
IV. Hierarchical Cluster Analysis (HCA)
These four algorithms have both positives and negatives. Out of
this K-Means algorithm is selected since it proved more effective
when tested against the test dataset.
Then the model is developed using K-means clustering
algorithm. K-means clustering is used to find groups which
have not been explicitly labeled in the dataset. This
algorithm can be used to confirm assumptions about what
types of groups exist or to identify unknown groups in
complex datasets. The algorithm works iteratively to assign
each data point to one of the K groups based on the features
that are provided. Data points are clustered based on feature
similarity. The results of the K-means clustering algorithm
are [5]:
I. The centroids of the K clusters, which can be used to
label new data.
II. Labels for the training data (each data point is assigned
to a single cluster)
The geometry/metric utilized in the K-means algorithm is
distances between points. The main conception is to determine
k centers, one for each cluster. So, number of clusters have to
be fixed here. There are four types of games available for
Figure 1. High-Level Diagram of the System
those four memory factors with three difficulty levels: easy,
Following section discuss the workflow of the proposed solution medium and hard. So, the number of clusters is specified as 12
in detail. (k=12).
The K-means algorithm is aimed to decide on centroids that
3.1 Developing an evaluation phase to test minimize the inertia, or within-cluster sum of squared criterion.
children to record their initial status 𝑛
A deep study is conducted to discover the psychological factors 2
∑ 𝑚𝑖𝑛⁡(‖𝑥𝑗 − 𝜇𝑖 ‖ )
behind child’s brain development. There are many factors that
𝑖=0
affect brain development. Mental stimulation, nutrition, physical
activity, interpersonal relationships, and genetics are some of
them. Families have an extremely important ongoing influence This algorithm partitions the dataset of N samples X into K
on children’s development. The community and service disjoint clusters C. 𝜇𝑖 ⁡denotes the mean of the clusters in the
environments in which children and families interact also play a sample.
key role in supporting optimal development[9]. As it is only
possible to stimulate mentally with the application, four factors The algorithm analyzes the data provided, looks for specific
are selected to evaluate in children which occupy a prominent patterns, and uses the results of this analysis to develop optimal

70
parameters for creating the model. Once the algorithm has been The milestones specific to ages that contain the details of the
run, and the groups are defined, any new data can be easily activities and skills are collected from a pediatrician. This is used
assigned to the correct group. as standards. This standard data is compared against the child data
collected in the Evaluation phase, and the results are produced.
The goal of the machine learning model is to suggest the most
relevant games specific to the child to develop the skill in Initially, when the child uses the system for the first time, the
which the child lags. The model analyses the data and child undergoes an Evaluation phase to determine the current
determines whether to direct the child to advanced levels in the brain growth level. The stated four factors are evaluated during
games or to provide proper training with easy/medium levels. the testing process. Based on the child’s result, his status is
determined.
3.3 Model training By analyzing the data specific to the child, the model redirects the
The model is refined through model training. Number of child to the suitable games and appropriate levels. Then the child
iterations and the richness of the data are raised. Data will be analyzed periodically, and the test results will be sent to
preprocessing is used to transform raw data into an parents through a report.
understandable format. Data preprocessing is done initially to
clean the dataset by removing unnecessary features and null
values. The process of reducing the number of random
4. RESULTS & DISCUSSION
variables under consideration, by obtaining a set of principal This section proofs that the accuracy of the model is increased
values is dimensionality reduction. Dimensionality reduction through model training, and thus the performance. The model is
is performed as having data of high dimensionality is refined through several ways to increase its performance. Earlier
problematic because it can mean high computational cost to when the accuracy is checked against test dataset, it was quite
perform learning and inference and it often leads to overfitting. lower. Then it is decided to train the model in several ways.
More instances were added to the dataset to increase the data
richness and to give training to the model by making the model
3.4 Model testing familiar with various combinations of scores. And also, the
Testing is making a prediction upon a set of data we already know improvement in the child’s mental development is discussed in
the result, and we check how good the algorithm predicted the this section.
desired data. From the results, it could be verified if the
algorithm has learned or not. Datasets and Preprocessing
Data collection became comparatively complex here. The dataset
Evaluating the performance of a clustering algorithm is not as is prepared by making children to play only the evaluation part of
trivial as counting the number of errors or the precision and recall the app and store it. Raw data cannot be used because it might be
of a supervised classification algorithm. noisy. So, data preprocessing is done. Data cleaning, data
K-fold cross-validation algorithm is selected to evaluate the integration, data transformation, and data reduction are done
model. The procedure has a single parameter called k that refers sequentially. In data cleaning, data is cleansed through processes
to the number of groups that a given data sample is to be split into. such as filling in missing values, smoothing noisy data or
When we assign a value for k, it may be used in place of k in resolving the inconsistencies in the data. Data with different
reference to the model, such as k=5 becoming 5-fold cross- representations are put together and conflicts within the data are
validation. resolved in data integration. Data is normalized, aggregated and
generalized in data transformation. Data reduction aims to
present a reduced representation of data. 80% of the dataset is
3.5 Gamification used as the test set and the remaining 20% as the training set. A
Gamification is aimed to develop the cognitive skills via games in sample of the dataset is shown here.
an interactive environment. The games are developed to test
memory power, concentration, understanding objects, shape
variance, and easy vocabulary to examine the brain growth level.
For each category, the games have been developed with different
difficulty levels. If the child finishes the level, then the system
proceeds to the next level. Suppose the child fails a level, the
system identifies the child’s interested areas and train the child
based on those areas.
Apparatus
The technical requirements of the research are as follows.
1. Android phone with operating system 4.1 or higher Android
platform is chosen because it is a widely used mobile
operating system.
2. Android studio - Android studio is the officially integrated
environment (IDE) for Google’s Android operating system. Figure 2. A sample of the dataset
3. Firebase
4. Windows 10 Selected K-Means algorithm has done a good job when proper
training is given to that. The accuracy of the model increased
Design and procedures from 73.10% to 88.49% after the training period. In model
With all these procedures and technical requirements, the research training as more instances of the dataset is added, the model could
that described in the paper is designed as follows:

71
learn more, and the experience of the model has been improved 7. REFERENCES
and thus the performance. [1] “Importance of early childhood development | Encyclopedia
When it comes to the system users, that is children in this project on Early Childhood Development.” [Online]. Available:
they too have improved in the skills in which they lagged earlier. http://www.child-encyclopedia.com/importance-early-
The very first time when a child is registered with the application childhood-development. [Accessed: 03-Aug-2018].
and logs in, he undergoes an Evaluation phase. The child does [2] D. Eddie and D. Schmid, “Quality Matters : A Policy Brief
not undergo the Evaluation phase often. Very first time when Series Brain Development and Early Learning Quality
using the app, the child is evaluated and the after one or two Matters : A Policy Brief Series,” Wisconsin Counc. Child.
months once the child is done with the training then it undergoes Fam., vol. 1, pp. 2–5, 2007.
to the Evaluation phase to re-evaluate himself.
[3] P. D, “Factors Affecting Early Childhood Growth and
For research analysis purpose, a child’s details are recorded and Development: Golden 1000 Days,” Adv. Pract. Nurs., vol.
made the child to use the application. The initial scores for each 01, no. 01, pp. 1–4, Nov. 2016.
category games are recorded. Then according to the model’s [4] “Developmental milestones record: MedlinePlus Medical
suggestion, the child is redirected to that particular game and Encyclopedia.” [Online]. Available:
level. Training is given to the child via gamification. After the https://medlineplus.gov/ency/article/002002.htm. [Accessed:
successful completion of the training, i.e. the child is re-evaluated 09-Aug-2018].
in the Evaluation phase. As expected, there is an improvement in
the score in which the child lagged. [5] “Introduction to K-means Clustering.” [Online]. Available:
https://www.datascience.com/blog/k-means-clustering.
Challenges [Accessed: 19-Aug-2018].
 Developing a model with higher accuracy is challenging. [6] “The Science And The Benefits of Gamification In
 Implementing RPG (Role Playing Game) mechanics and eLearning - eLearning Industry.” [Online]. Available:
making it interesting to motivate user interaction. https://elearningindustry.com/science-benefits-gamification-
elearning. [Accessed: 09-Aug-2018].
 Developing minigames inside the RPG platform to improve
the brain skills. [7] “Success in Gamification of the Workplace - business.com.”
[Online]. Available: https://www.business.com/articles/5-
Approach statistics-that-prove-gamification-is-the-future-of-the-
As the application is for children, it has been decided to workplace/. [Accessed: 09-Aug-2018].
implement a role-playing game (RPG) mechanics to develop the [8] “Gamification Research Results: Gamification Neuroscience
game with the use of gamification technique. To Reduce Decision Error And Build A Learning
Community - eLearning Industry.” [Online]. Available:
5. CONCLUSION https://elearningindustry.com/gamification-research-results-
This research paper discussed on the research project ‘Smart gamification-neuroscience-reduce-decision-error-build-
Monitor for Tracking Child’s Brain Development’ A way to track learning-community. [Accessed: 09-Aug-2018].
and develop a child’s brain is proposed in this work using [9] Y. Ahuja, “Australian early developmetn census,” Factors
Machine Learning and Gamification technologies. Influ. Brain Dev. Child., no. Figure 1, 2014.
An Android application is developed which is capable of tracking
a child’s brain growth by evaluating the child and improving it
via gamification. The proper ways to test brain growth factors are
discovered in an Evaluation phase through proper tests. By
analyzing the collected data, the model suggests suitable games.
And the user is directed to that particular game and level. Then
training is given to the child via gamification.
The system interacts with the child and trains it accordingly to
increase his brain growth through gamification. The system also
generates a report and notifies parents, so that parents can guide
and interact with the child in a way the child needs.
The machine learning model is refined in several ways to improve
its accuracy. For this model training is done. And also, it is very
important to test and evaluate the model to find out how
accurately the model predicts. The objective is to evaluate user,
trains model and evaluate the model which are accomplished.

6. FUTURE WORKS
It is planned to introduce the application in cross-platform with
more new features like new gaming area, rewards, quests,
characters, etc. The model will be refined through several ways
for higher accuracy. A new factor to improve the cognitive skills
will be added later.

72

Vous aimerez peut-être aussi