sensors

Article
Smart Interactive Education System Based on
Wearable Devices
Jia-Ming Liang 1,2 , Wei-Cheng Su 1 , Yu-Lin Chen 1 , Shih-Lin Wu 1,3,4,5 and Jen-Jee Chen 6,7, *
1 Department of Computer Science and Information Engineering, Chang Gung University,
Taoyuan 33302, Taiwan
2 Department of General Medicine, Chang Gung Memorial Hospital, Taoyuan 33378, Taiwan
3 Department of AI Innovation Research Center, Chang Gung University, Taoyuan 33302, Taiwan
4 Department of Cardiology, Chang Gung Memorial Hospital, Taoyuan 33305, Taiwan
5 Department of Electrical Engineering, Ming Chi University of Technology, New Taipei City 24301, Taiwan
6 Department of Electrical Engineering, National University of Tainan, Tainan 70005, Taiwan
7 College of Artificial Intelligence, National Chiao Tung University, Tainan 71150, Taiwan
* Correspondence: jjchen@mail.nutn.edu.tw; Tel.: +886-606123 (ext. 7799)




Received: 15 June 2019; Accepted: 17 July 2019; Published: 24 July 2019


Abstract: Due to the popularity of smart devices, traditional one-way teaching methods might not
deeply attract school students’ attention, especially for the junior high school students, elementary
school students, or even younger students, which is a critical issue for educators. Therefore, we
develop an intelligent interactive education system, which leverages wearable devices (smart watches)
to accurately capture hand gestures of school students and respond instantly to teachers so as to
increase the interaction and attraction of school students in class. In addition, through multiple
physical information of school students from the smart watch, it can find out the crux points of the
learning process according to the deep data analysis. In this way, it can provide teachers to make
instant adjustments and suggest school students to achieve multi-learning and innovative thinking.
The system is mainly composed of three components: (1) smart interactive watch; (2) teacher-side
smart application (App); and (3) cloud-based analysis system. Specifically, the smart interactive
watch is responsible for detecting the physical information and interaction results of school students,
and then giving feedback to the teachers. The teacher-side app will provide real-time learning
suggestions to adjust the teaching pace to avoid learning disability. The cloud-based analysis
system provides intelligent learning advices, academic performance prediction and anomaly learning
detection. Through field trials, our system has been verified that can potentially enhance teaching
and learning processes for both educators and school students.

Keywords: intelligent interactivity; instant feedback; wearable device; data analysis; learning
concentration; education system

1. Introduction
Smart devices have brought tremendous changes to human life, especially for the young people.
Traditional teaching skills might be unable to attract school students’ attention nowadays and a new
intelligent teaching system is required for the modern society. According to this, we investigate a smart
interactive educational learning system based on wearable devices, which integrates smart watches with
a smartphone application (App) that can interact with school students and educators, and make
teaching and learning processes more interesting and concentrated. This system leverages the machine
learning technique and data analysis based on school students’ interaction data, physical information,
and learning performance, and then models them as the learning benefit relevance. In this way,
the teacher can easily control the learning situation in the classroom and also can adjust the teaching

Sensors 2019, 19, 3260; doi:10.3390/s19153260 www.mdpi.com/journal/sensors

Sensors 2019, 19, 3260 2 of 8

contents appropriately in time to prevent school students from distracted in the class. Specifically,
the system can predict the trend of learning performance by analyzing the interaction feedback,
physical information, and test scores, and provide suitable learning suggestions to improve teaching
and learning efficiency.
Sensors 2019, 19, x FOR PEER REVIEW 2 of 8
The major contributions of this paper are three-fold. First, this is the first work to integrate
Specifically,
with smart watch, thesmartphone,
system can predict the could-based
and the trend of learning performance
service by analyzing
to enhance the interaction
the teaching and learning
feedback, physical information, and test scores, and provide suitable
processes for both educators and students. Second, the system is designed based learning suggestions to improve
on wearable
teaching and learning efficiency.
sensing technologies and machine learning techniques so as to provide interesting interactive functions,
The major contributions of this paper are three-fold. First, this is the first work to integrate with
including: (1) random grouping; (2) hand-gesture answer; (3) group competition; (4) instant questioning;
smart watch, smartphone, and the could-based service to enhance the teaching and learning
(5) automatic
processes roll
forcall;
bothand (6) analysis
educators and prediction
and students. of learning
Second, the performance,
system is designed basedwhich can make
on wearable students
sensing
more concentrated
technologies and in machine
class and enjoy learning.
learning techniques Third,
so as tothrough
providethe field trials,
interesting our system
interactive has been
functions,
validated at Taoyuan
including: Municipal
(1) random JieShou
grouping; (2) Junior High School,
hand-gesture answer;Taiwan.
(3) group competition; (4) instant
questioning; (5) automatic roll call; and (6) analysis and prediction of learning performance, which
2. Related Work
can make students more concentrated in class and enjoy learning. Third, through the field trials, our
Insystem has been validated
the literature, at Taoyuanhave
several researches Municipal
studied JieShou Junior High
the classroom School, Taiwan.
interactive feedback systems [1–11].
The work [1] explores the influence of the traditional push-button voting system on learning
2. Related Work
concentration. It points out that the interactive feedback system can help to increase school students’
concentration;In the however,
literature, several researches
it neglects have studied
the features the classroom
to inspire interactive
students. feedback
In [2,3], thesystems
authors [1–design
11]. The work [1] explores the influence of the traditional push-button voting
a smartphone application to attract students’ attention; however, they may distract students by other system on learning
concentration. It points out that the interactive feedback system can help to increase school students’
gaming applications from smartphones. In [4], the authors use short messages on mobile phones to
concentration; however, it neglects the features to inspire students. In [2,3], the authors design a
increase the interactivity in class, but it may not prevent students from repeatedly asking the same
smartphone application to attract students’ attention; however, they may distract students by other
questions
gamingandapplications
discussing from
othersmartphones.
things irrelevantIn [4],to
thethe class. use
authors In [5],
shortthe authorson
messages propose
mobile using
phonesa to‘clicker’
deviceincrease
in the classroom to improve
the interactivity the
in class, butinteraction
it may notbetween teachersfrom
prevent students andrepeatedly
students. asking
However, this device
the same
only can recordand
questions the discussing
students’ other
attendance
things and answer
irrelevant simple
to the class.questions;
In [5], the thus, it cannot
authors proposeprovide
using a better
‘clicker’
interactive device in such
functions the classroom to improveanswer,
as hand-gesture the interaction
random between teachers
grouping, and students.
group However,
competition, analysis,
this device only can record the students’ attendance and answer simple
and prediction of learning performance. In [6], the authors proposes a web-based classroom response questions; thus, it cannot
provide better interactive functions such as hand-gesture answer, random grouping, group
system to promote active learning. However, it may not prevent students from playing other web
competition, analysis, and prediction of learning performance. In [6], the authors proposes a web-
gamesbasedin the classroom. In [7,8], the authors propose a game-based student response system (SRS)
classroom response system to promote active learning. However, it may not prevent students
to improve students’
from playing participation;
other web games inhowever, these
the classroom. Ingames may
[7,8], the alsopropose
authors lead to adistraction
game-basedfrom learning.
student
From response
[9], the authors
system (SRS) to improve students’ participation; however, these games may also lead to[10,11],
show that the exercise and concentration are highly correlated. In
the authors mention
distraction from that the wearable
learning. From [9], the devices
authorsareshow
emerging
that thetoexercise
enhance andlearning efficiency.
concentration Therefore,
are highly
based correlated.
on the above In [10,11], the authorsitmention
observations, motivates that us
thetowearable devices
investigate are emerging
a smart to enhance
interactive learningsystem
education
efficiency.devices
with wearable Therefore, based on
to achieve anthe above observations,
innovative educating and it motivates
learning.us to investigate a smart
interactive education system with wearable devices to achieve an innovative educating and learning.
3. System Design
3. System Design
Figure 1 shows the architecture of our system, which consists of three components: (1) smart
interactiveFigure
watch;1 shows the architecture
(2) teacher-side smartofapp;
our system,
and (3) which consistsanalysis
cloud-based of three components:
system. (1) smart

WIFI Cloud
Server

Smartphone
Upload/Download
Teacher
Students’ data
Bluetooth

Smart Smart
watch watch Smart
watch

Figure 1. System architecture.

Figure 1. System architecture.
 26BD ⚽ \textSoccerBall SOCCER BALL

26C5 ⛅ \textSunCload SUN BEHIND CLOUD

26C6 ⛆ \textRain RAIN

26D4 ⛔ \textnoway NO ENTRY

26F0 ⛰ \textMountain MOUNTAIN

Sensors 2019, 19, 3260

26FA ⛺ \textTent TENT 3 of 8
2701 ✁ \textScissorRightBrokenBottom UPPER BLADE SCISSORS

2702 ✂ \textScissorRight BLACK SCISSORS

1. Smart Interactive ✃ to collect school students’ learning interactions and
Watch:\textScissorRightBrokenTop physical information,
Sensors 2019, 2703
19, x FOR PEER REVIEW LOWER BLADE SCISSORS
3 of 8
where learning
2704 interactions
✄ include the number
\textScissorHollowRight of times for raising hands and answering
WHITE SCISSORS questions,
1.andSmart
the 2706
corresponding
Interactive ✆ Watch: response time in
\textPhoneHandset
to collect personal/group
school students’ learning competitions;
interactions physical
and
TELEPHONE information
physical
LOCATION SIGN

includes 2707
information,
heart rate, ✇ exercise
where \textTape
learning interactions
strength (number include the number
of walking of times
steps), activityfor frequency,
raising
TAPE DRIVE hands etc.and It is also
answering
responsible report✈ such
2708to questions, \textPlane
and the corresponding
information response time
to the teacher-side in personal/group
application AIRPLANE
and cloud-basedcompetitions; system for
physical2709information✉ \textEnvelope
includes heart rate, exercise strength (number of walking steps), activity
ENVELOPE
further analyses.
270C etc. It
frequency, ✌ is also\textPeace
responsible to report such information to the teacher-side VICTORY HANDapplication
2. Teacher-side 270D
Smart to receive the interactive information from the smart
App:\textWritingHand
✍system watch and provide
and cloud-based for further analyses. WRITING HAND
special functions
270E
2. Teacher-side ✎for teaching
Smart App:
so that the teacher can immediately understand
\textPencilRightDown
to receive the interactive information from theLOWER smartRIGHTschool
watch
PENCIL and students’
learning status
270F in ✏the classroom. The
\textPencilRight specific functions are as
provide special functions for teaching so that the teacher can immediately understand school follows: PENCIL

2710 ✐ status\textPencilRightUp
a. students’
Randomlearning in thestudents
classroom. Thebespecific functionsand are as follows:UPPER RIGHT PENCIL
grouping: school will informed divided into the specific number
2711 ✑ \textNibRight WHITE NIB
a.of groups
Random
2712
according
grouping:
✒ to the
school teacher’s
students
\textNibSolidRight will be setting,
informed such andasdivided
the group size,
into the gender,
specific
BLACK NIB
number or learning
of
performance,
groups
2713 ✓etc. This
according can teacher’s
to the save lots setting,
\textCheckmark of classsuchtimeas andtheavoid
groupthe unfair
size, doubts.
gender,
CHECK MARK
or learning
performance,
b. Hand-gesture
2714 ✔ etc. This
answer: can save lots
to recognize
\textCheckmarkBold theofanswer
class time andyes-no
of the avoid the unfair through
question doubts.
HEAVY CHECKtheMARK
hand-gesture
b.of Hand-gesture
2715 ✕ answer: to recognize the answer
‘#’ and ‘ ’ from smart watches (by moving hands in a circle and a cross
\textXSolid of the yes-no question through
MULTIPLICATION
theX
hand- motion,
gesture
2716 of ‘○’
✖ and \textXSolidBold
‘╳’ from smart watches (by moving hands in a circle and aMULTIPLICATION
cross motion,
respectively) and detect the answer of the multiple-choice question through the hand-gesture HEAVY X
respectively) and \textXSolidBrush
detect the answer of the multiple-choice question through the hand-
of ‘↑’, ‘↓’, ‘←’,✗and ‘→’
2717 by waving hands in up, down, left, and right directions. BALLOT X
After receiving
gesture
2719 of ‘↑’,
✙ ‘↓’,\textPlusOutline
‘←’, and ‘→’ by waving hands in up, down, left, and right directions.
OUTLINED GREEK CROSS
theAfter
answers from the smart watches fromthrough Bluetooth communication, this app will check this the
271A receiving ✚ answers
\textPlus smart watches through Bluetooth communication,
HEAVY GREEK CROSS
results
app automatically
271Bwill check
and send
✛ the\textPlusThinCenterOpen the feedback to the smart
results automatically and send the feedback to the smart watches and then perform
watches
OPEN CENTRE CROSS
vibration.
and
Note that
then
271C the✜
perform hand-gesture
vibration. activities
Note that the
\textPlusCenterOpen are recognized activities
hand-gesture based onare therecognized
decision tree
based
HEAVY OPEN
scheme,
CENTRE on thewhich
CROSS

is adecision
machine
271D ✝learning
tree scheme, model,
which isby
\textCross referring
a machine the values
learning model, ofbythree-axis
referring accelerometer
the
LATINvalues
CROSS of three-from the
axis
smart271E accelerometer
watch. from theofsmart
✞The flowchart watch. The
hand-gesture
\textCrossOpenShadow flowchart of
recognition hand-gesture
is shown inSHADOWEDrecognition
Figure 2. LATIN CROSS
WHITE is
c. shown
271F in Figure
✟ 2. \textCrossOutline
Group competition: after grouping, the teachers can obtain the interaction results and rank of OUTLINED LATIN CROSS

scores to✠ achieve

c. Group2720 competition: after grouping, the teachers can obtain the interaction MALTESE
\textCrossMaltese resultsCROSS
and rank of
their group competition.
2721 scores✡to achieve
their group competition.
\textDavidStar STAR OF DAVID
d. d.Instant questioning: school students can ask questions quietly and privately by clicking on
Instant ✢
2722 questioning: school students can ask questions quietly and privately
\textFourAsterisk by clicking ASTERISK
FOUR TEARDROP-SPOKED on
thethe
smart
2723 watch,
✣
smart watch, which
which is more suitable for introverted school students.
is more suitable for introverted school students.FOUR BALLOON-SPOKED ASTERISK
\textJackStar
e. e.Automatic
Automatic
2724 roll
✤ call:
roll call:through
through the the connection
\textJackStarBold connectionbetween between thethe smart
smart watch,
watch, the
HEAVY the
roll rollcan
FOURcall callbecan
BALLOON-SPOKED be
ASTERISK

realized ✥
2725 automatically
realized automatically to
toreduce
reduce the
\textClowerTips thewaste
wasteofoftime time in in class.
class. FOUR CLUB-SPOKED ASTERISK

2726 ✦ \textFourStar BLACK FOUR POINTED STAR

2727 ✧ \textFourStarOpen WHITE FOUR POINTED STAR

272A ✪ \textFiveStarOpenCircled
G-sensor CIRCLED WHITE STAR

272B ✫ \textFiveStarCenterOpen
Operation OPEN CENTRE BLACK STAR

272C ✬ X-axis
\textFiveStarOpenDotted SUM MAX
DecisionBLACK CENTRE WHITE STAR
272D ✭ \textFiveStarOutline Tree OUTLINED BLACK STAR
Y-axis MIN tan-1
272ESmart ✮ \textFiveStarOutlineHeavy HEAVY OUTLINED BLACK STAR

272Fwatch ✯ Z-axis
\textFiveStarConvex Integral PINWHEEL STAR

2730 ✰ \textFiveStarShadow SHADOWED WHITE STAR

2731 ✱ \textAsteriskBold HEAVY ASTERISK

2732 ✲ \textAsteriskCenterOpen OPEN CENTRE ASTERISK

2734 ✴ Figure2.2.Flowchart
Figure Flowchart of
\textEightStarTaper
ofhand-gesture
hand-gesture recognition.
recognition. EIGHT POINTED BLACK STAR

2735 ✵ \textEightStarConvex EIGHT POINTED PINWHEEL STAR

3. 3.Cloud-based
Cloud-based Analysis System: through
AnalysisSystem: through the thecollected
collected datadatafromfrom smart
smart watches
watches and teacher-side
and teacher-side
app,app,thethe
cloudcloud system
systemcan candeeply
deeply analyze
analyze and andpredict
predictthethe learning
learning performance
performance basedbased on theon the
machine
machine learning
learning technology
technology by referring
by referring the results
the results of school of school
students’ students’
answer, answer,
group group
competition,
46
competition, instant feedback, physical information, and then
instant feedback, physical information, and then conduct prediction for future learning scores. conduct prediction for future
learning scores.
In addition, the system In addition,
can provide the system
learningcan provide
advices learning
and teaching advices and teaching
suggestions forsuggestions
school students
for school students and teachers, respectively. Specifically, we leverage two prediction methods
and teachers, respectively. Specifically, we leverage two prediction methods with the analysis
with the analysis regression of score prediction, called Linear Regression (LR) and Local Weights
regression of score prediction, called Linear Regression (LR) and Local Weights Linear Regression
Linear Regression (LWLR) [12]. The main difference between the two prediction methods is that
(LWLR) [12]. The
LR calculates only main difference
a single weight value between for allthe
thetwoschool prediction
students in methods
class andisLWLR that finds
LR calculates
the
onlyweight
a single weight
value value forthat
individually all the school
matches students
each student’s in class
resultand LWLR finds
to achieve the weight
an accurate score value
individually
prediction.that matches each
In addition, eachestablished
student’s result modeltowill achieve
adjustan theaccurate
weightsscore
after aprediction.
period of time In addition,
to
each established model will adjust the weights after a period of time to make the predicted score
interval more realistic. The details of the prediction method are described as follows:
Sensors 2019, 19, 3260 4 of 8

• Method I: Linear Regression (LR)

For the collect information from class interaction, it calculates the test scores, the overall class
situation, and weight of the all class, and then takes the weight value into the following functions and
constructs a model:
y = θ0 + θ1 X1 + θ2 X2 + θ3 X3 + θ4 X4 (1)

y = nθ0 + θ1 x1 + θ2 x2 + θ3 x3 + θ4 x4
 P P P P P



x1 y = θ0 x1 + θ1 x21 + θ2 x1 x2 + θ3 x1 x3 + θ4 x1 x4

 P P P P P P




x2 y = θ0 x2 + θ1 x1 x2 + θ2 x22 + θ3 x2 x3 + θ4 x2 x4

 P P P P P P

 (2)

x3 y = θ0 x3 + θ1 x1 x3 + θ2 x2 x3 + θ3 x3 + θ4 x3 x4

 P P P P P 2 P




 P x y = θ P x + θ P x x + θ P x x + θ P x x + θ P x2

4 0 4 1 1 4 2 2 4 3 3 4 4 4
−1
θ̂ = (XT X) XT Y (3)

where y is the test score, x1 is the instant question score, x2 is the hand-gesture answer, x3 is group
competition result, and x4 is the amount of personal exercises. This can calculate the most suitable
coefficients in the class. Finally, by using the matrix operation in Equation (2), the weight values are
calculated as a regression model.
• Method 2: Local Weights Linear Regression (LWLR)
When calculating the predicted scores, unreasonable score predictions may occur when the general
performance of the all class is very good or very bad, leading to difficulty in detecting the situation
that some school students are more prominent or whose performance needs improvement. In order to
reduce the inaccuracy of these predictions, we use the local weighted linear regression to give each
student their own weights; then, more accurate prediction of future scores will be:

2
J (θ) = Σi w(i) ( y(i) − θT x(i) ) (4)
  2 
 χ(i) − x 
w(i) = exp−
 
2
 (5)
 2k 

and:  −1
θ̂ = XT WX XT WY (6)

where Equation (4) means LR plus the respective weight which yields the regression result of LWLR.
Equation (5) is the exponential decay function, which can adjust the value of k to find the best predicted
model. Finally, through the matrix operation in Equation (6), it can find the coefficients that fit each
student and then construct an individual prediction score model.

4. System Implementation and Performance Verification

4.1. System Implementation

The system has been implemented. The hardware specifications are listed as follows. The smart
watch is developed based on Alfa Bracelet DS62 (produced by Alfaloop Inc., Taiwan, as shown in
Figure 3a), where the operating system is Alfa OS v1.0. Note that the wireless module is Bluetooth 4.0,
which is used for connecting with the teacher-side smart app. The smart app is developed based on
Android 6.0.1 and operates on the smartphone (model: ze550kl, produced by ASUSTeK Computer
Inc., Taiwan, with Qualcomm S615, 2 GB RAM and 16 GB ROM). The cloud system operates on the
platform of Intel Core i5-6400 CPU 2.70 GHz, DDR3-1600 4 GB SDRAM, 1 TB HDD, and the operating
system is Windows 10 Pro (64-bit), where the database is MariaDB v10.
Sensors 2019, 19, x FOR PEER REVIEW 5 of 8

Sensors 2019, 19, 3260 Specification Content 5 of 8

Sensors 2019, 19, x FOR PEER REVIEW CPU Nordic52832 5 of 8
Wireless Connection Bluetooth - BT 4.2
Motion Sensor 3D G-sensor (KX023)
Specification Content
Heartrate Monitor PAH8001
CPU Nordic52832
Flash Font library IC
Wireless Connection Bluetooth - BT 4.2
Display 0.89” OLED (128*32)
Motion Sensor 3D G-sensor (KX023)
Battery
Heartrate Monitor 3.7V / 90 mAh
PAH8001
Interactive
Flash components One
Fonttouch
librarybutton
IC & vibrator
Debug Interface
Display external 2 GPIO
0.89” OLED (UART) & BLE
(128*32)
Battery 3.7V / 90 mAh
(a) Interactive components
(b)One touch button & vibrator
Debug Interface external 2 GPIO (UART) & BLE
Figure 3. The
(a) smart specification
watch device of the Alfa Bracelet DS62.
(b) hardware specification

4.2. Performance Verification Figure 3. The specification of the Alfa Bracelet DS62.
The following
Performanceresults
4.2. Performance show the field experiments at Taoyuan Municipal JieShou Junior High
Verification
4.2. Verification
School, Taiwan. The total number of students was 18. In the classroom, the teacher divided all the
The following
The following results show
results show thethe field
field experiments
experiments at Taoyuan
at Taoyuan Municipal
Municipal JieShou Junior High
students into six groups through the function of random grouping, where JieShou
each group Juniorhas Highthree
School,
School, Taiwan.
Taiwan. The
The total
total number
number of
of students
students was
was 18.
18. In
In the
the classroom,
classroom, the
the teacher
teacher divided
divided all
all the
the
students. Then, the teacher asked ten questions by the function of instant questioning and five
students into
students intosix
sixgroups
groups through
through the the
function of random
function grouping,
of random where where
grouping, each group
each has threehas
group students.
three
questions
Then,
were performed by the group competition. Finally, three tests were completed by hand-
students. Then, the teacher asked ten questions by the function of instant questioning andwere
the teacher asked ten questions by the function of instant questioning and five questions five
gesture answering.
performed by the group competition. Finally, three testsFinally,
were completed
questions were performed by the group competition. three testsby hand-gesture
were completedanswering.
by hand-
For theFor verification
the of learning
verification of process,
learning we we
process, conduct
conduct thetheexperiment
experiment totoverify
verify the
theimpact
impact of of the
gesture answering.
students’ interactive
the students’ results
interactive of
For the verification
from
results the proposed
fromprocess,
learning the proposed system
system
we conduct
on the learning
theonexperiment
the learning performance,
to performance,
verify the impact
where
where of the
the
the
interactive results
interactive
students’ are are
results
interactive calculated
calculated
results based
from the on
based onthe
thehand-gesture
proposed hand-gesture
system on the answer,
answer, group
group
learning competition,
competition,
performance, where instant
instant
the
questioning,
questioning, etc., etc., collected
collected by by
the the proposed
proposed system.
system. InIn Figure
Figure 4,
4, we
we can
can see
see
interactive results are calculated based on the hand-gesture answer, group competition, instant that
that most
most of the
of students’
the students’
learning
learning performances
performances
questioning, are are positively
positively
etc., collected by correlated
correlated
the proposed totothe
system. the interactive
Ininteractive
Figure 4, we results.
can seeThis
results. thatismost
This isbecause
because our
of the system
our system
students’
can
can attract attract
learning students’
students’
performances attention,
attention, making them
makingcorrelated
are positively concentrate
them concentrate on the
on the
to the interactive teaching
teaching
results. content
This is content and
becauseand enjoy
enjoy
our system the the
learning
can process.
attract students’ attention, making them concentrate on the teaching content and enjoy the
learning process.
learning process.
100 350
100 Final Score 350
90 Final Score
90 Interactive Score 300
Interactive Score 300
80
80
Learning Performance

250
Learning Performance

70
Interactive Results

70 250
Interactive Results

60 60 200
200
50 50
40
150
150
40
30 30 100
100
20 20
50
10 50
10
0 0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Student ID
Student ID
Figure 4. Comparison on the impact of interactive results on learning performance.
Figure 4. Comparison on the impact of interactive results on learning performance.
Figure 4. Comparison on the impact of interactive results on learning performance.
For the analysis
For the analysis and
and prediction
prediction of
of school
school students’
students’ grades,
grades, Figures
Figures 55 and
and 66 show
show the
the results
results
between
For the
the predicted
the analysis
between and
and actual
actual scores
and prediction
predicted by
of school
scores LRstudents’
by LR and LWLR,
and LWLR, respectively.
grades, Note
FiguresNote
respectively. that6we
5 and
that we use multiple
show
use multiple
the results
between the predicted and actual scores by LR and LWLR, respectively. Note that we use group
scores in the classroom, including the results of personal/group answers, hand-gesture results, multiple
scorescompetitions, and amount
in the classroom, of exercises,
including testof
the results scores, and exam scores,
personal/group to construct
answers, the models,
hand-gesture where
results, group
competitions, and amount of exercises, test scores, and exam scores, to construct the models, where
 Sensors 2019, 19, 3260 6 of 8

Sensors 2019, 19, x FOR PEER REVIEW 6 of 8

scores in the classroom, including the results of personal/group answers, hand-gesture results, group
Sensors 2019, 19, x FOR PEER REVIEW 6 of 8
competitions, and amount of exercises, test scores, and exam scores, to construct the models, where 2/3
2/3 of the data is used for the modeling and 1/3 of the data is used for the testing. Then, after
of the data is used for the modeling and 1/3 of the data is used for the testing. Then, after exchanging
2/3 of the data
exchanging theistest
used
dataforand
themodeling
modelingdata,andtwo1/3 more
of theaccurate
data is models
used forarethe testing. Finally,
obtained. Then, after
parts
the test data and modeling data, two more accurate models are obtained. Finally, parts of the scores
exchanging
of the scoresthe test
in the data and modeling data, two more accurate models are obtained. Finally, parts
in the course are course
applied are applied
for the modelforto the model
obtain to obtainscores.
the predicted the predicted
As can be scores. As average
seen, the can be seen,
ofthe
theaverage
scores in the course
difference are
between applied
the for the
predicted model
score to
andobtain
actual the predicted
score is less scores.
than As
9.6%,
difference between the predicted score and actual score is less than 9.6%, implying that the performance can be seen,
implying that
the average difference
the performance results
results are fairly between the
are fairly
consistent. predicted
consistent.
In addition, Figurescore and
In 7addition,actual
shows theFigure score
results7of is
shows less than
thescores
different 9.6%,
results implying
of different
intervals that
on thescores
the performance
intervals on the
predicted results
predicted
score are
differencesfairly
score consistent.
differences
by LR and LWLR.byIn addition,
WeLR and
can Figure
seeLWLR. 7
that the We shows the
can see
predicted results
thatofthe
score of different
thepredicted scores
score
interval 0~5 is of
intervals
moreon the predicted
accurate for both score
LR anddifferences by LR and LWLR. We can see that the predicted score of
LWLR methods.
100 actual score
100 actual score
90 predicted score
90 predicted score
80
80
70
70
60
Score

60
50
Score

50
40
40
30
30
20
20
10
10
0
0 1 2 3 4 5 6 78 9 10 11 12 13 14 15 16 17 18
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Student ID
Student ID
Figure 5. Comparison on predicted score and actual score of LR.
Figure 5. Comparison on predicted score and actual score of LR.
Figure 5. Comparison on predicted score and actual score of LR.

100 actual score

100 actual score
90 predicted score
90 predicted score
80
80
70
70
60
Score

60
Score

50
50
40
40
30
30
20
20
10
10
0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Student ID
Student ID
Figure 6. Comparison on predicted score and actual score of LWLR.
Figure 6. Comparison on predicted score and actual score of LWLR.
Figure 6. Comparison on predicted score and actual score of LWLR.
SensorsSensors
2019, 19, x FOR
2019, PEER REVIEW
19, 3260 7 of78 of 8

10 LR
9 LWLR
8
7
Score Difference

6
5
4
3
2
1
0
0~5 5~10 10~15 15~20 20~25

Score Interval

Figure 7. The results of predicted score differences between LR and LWLR.

Figure 7. The results of predicted score differences between LR and LWLR.
5. Conclusions
5. Conclusions
In this paper, we have introduced a smart interactive educational system which integrates with
thethis
In smart watch
paper, weand smartphone
have introduced appathat
smartcaninteractive
create moreeducational
interactions between teachers
system which and schoolwith
integrates
students in class while providing more innovative learning possibility. In addition, it can also inspire
the smart watch and smartphone app that can create more interactions between teachers and school
school students and increase their concentration in the classroom. The system leveraged machine
students in class while providing more innovative learning possibility. In addition, it can also inspire
learning technology and data analysis based on the interaction results, physiological information and
schoolstudy
students and increase
performance their concentration
of the students in made
in real-time. This the classroom. Theadjust
teachers easily systemtheleveraged machine
teaching process
learning
andtechnology
enhance theand dataefficiency
learning analysis based on theByinteraction
of students. field trials, results,
we havephysiological information
verified that this system canand
studypotentially
performance of the students in real-time. This made teachers easily adjust the
enhance teaching and learning process for educators and school students. In addition, teaching process
and enhance
it also can the learning
predict efficiency
the school of students.
students’ By field
learning scores trials,
with a verywe have
small verified In
difference. that
thethis system
future work,can
potentially
we willenhance teaching
keep studying andcreative
on the learning process
design for educators
of interactive andsystems
education school to
students. In addition,
further improve the it
learning
also can predictprocess for both
the school educatorslearning
students’ and students.
scores with a very small difference. In the future work,
we will keep studying on the creative design of interactive education systems to further improve the
Author Contributions: W.-C.S. and Y.-L.C. implemented the system and designed experiments; J.-M.L., S.-L.W.,
learning
and process forand
J.-J.C. wrote both educators
revised and students.
the paper.
Funding: This research is co-sponsored by MOST 106-2221-E-182-015-MY3, 106-2221-E-182-013-MY3,
Author Contributions: W.-C.S.108-2218-E-006-044,
107-2221-E-024-001-MY3, and Y.-L.C. implemented the system and designed
108-2813-C-182-002-E, ITRI, andexperiments;
Chang GungJ.-M.L., S.-L.W.,
Memorial
Hospital,
and J.-J.C. Taoyuan.
wrote and revised the paper.
Acknowledgments: Thanks B.-J.W. and P.-Y.H. for their helps and consultations.
Funding: This research is co-sponsored by MOST 106-2221-E-182-015-MY3, 106-2221-E-182-013-MY3, 107-2221-
Interest: The authors108-2813-C-182-002-E,
Conflicts of108-2218-E-006-044,
E-024-001-MY3, declare no conflict of interest.
ITRI, and Chang Gung Memorial Hospital, Taoyuan.

Acknowledgments:
References Thanks B.-J.W. and P.-Y.H. for their helps and consultations.
Conflicts
1. of Interest:
Lin, The authors
K.-W. Study declare
on the impact no conflict
of instant of interest.
feedback mechanism on school students’ learning concentration.
Sci. Educ. J. 2014, 22, 87–107.
References
2. Song, S.-J. The Study of Implementation and Assessment of Multimedia Mobile Classroom Response System.
Master’s Thesis, Information Management Institute, Kaohsiung First University of Science and Technology,
1. Lin, K.-W. StudyTaiwan,
Kaohsiung, on the impact of instant
2007. Available feedback
online: mechanism on school students’
https://ndltd.ncl.edu.tw/r/bq8urg learning
(accessed concentration.
on 3 July 2007).
Sci.
3. Educ. J. 2014, 22, 87–107.
Coca, D.M. Software socrative and smartphones as tools for implementation of basic processes of active
2. Song,physics
S.-J. The Study
learning of Implementation
in classroom: and Assessment
An initial feasibility study withof Multimedia
prospective Mobile
teachers. Eur. Classroom
J. Phys. Educ.Response
2017,
System. Master’s Thesis, Information Management Institute, Kaohsiung First University of Science and
4, 17–24.
4. Markett,Kaohsiung,
Technology, C.; Sa’nchez, Taiwan,
I.A.; Weber, S.; Tangney,
2007. B. Using
Available short
online: message service to encourage interactivity
https://ndltd.ncl.edu.tw/r/bq8urg in on
(accessed
the classroom.
July 3, 2007). Comput. Educ. 2006, 46, 280–293.
3. Coca, D.M. Software socrative and smartphones as tools for implementation of basic processes of active
physics learning in classroom: An initial feasibility study with prospective teachers. Eur. J. Phys. Educ. 2017,
4, 17–24.
4. Markett, C.; Sa’nchez, I.A.; Weber, S.; Tangney, B. Using short message service to encourage interactivity
in the classroom. Comput. Educ. 2006, 46, 280–293.
Sensors 2019, 19, 3260 8 of 8

5. Rana, N.P.; Dwivedi, Y.K. Using clickers in a large business class: Examining use behavior and satisfaction.
J. Mark. Educ. 2016, 38, 47–64.
6. Shea, K.M. Beyond clickers, next generation classroom response systems for organic chemistry. J. Chem. Educ.
2016, 93, 971–974.
7. Wang, A.I. The wear out effect of a game-based student response system. Comput. Educ. 2015, 82, 217–227.
8. Abidin, H.Z.; Zaman, F.H.K. Students’ perceptions on game-based classroom response system in a computer
programming course. In Proceedings of the IEEE International Conference on Engineering Education
(ICEED), Kanazawa, Japan, 9–10 November 2017; pp. 254–259.
9. Yang, H.-L. The Effect of Exercise Strength for Attention and Simple Reaction Time—Using Jump Rope
for Example. Master’s Thesis, National Pingtung University of Education, Taipei, Taiwan, 2014. Available
online: http://140.127.82.166/retrieve/19231/102NPTT0567014-001.pdf (accessed on 10 January 2014).
10. Liu, Z.-M. Wearable Devices: Innovative Business Models Research. Master’s Thesis, Shih Chien University,
Taipei, Taiwan, 2017. Available online: https://hdl.handle.net/11296/pt67tw (accessed on 22 December 2017).
11. Lu, Y.; Zhang, S.; Zhang, Z.; Xiao, W.; Yu, S. A framework for learning analytics using commodity wearable
devices. Sensors 2017, 17, 1382.
12. Harrell, J.; Frank, E. Regression Modeling Strategies: With Applications to Linear Models, Logistic and Ordinal
Regression, and Survival Analysis; Springer: Berlin/Heidelberg, Germany, 2015.

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).