0% found this document useful (0 votes)
10 views9 pages

1 - Human Sitting Postures Classification Based On Angular Features With Fuzzy-Logic Labeling

The study presents a system for classifying human sitting postures using angular features and fuzzy-logic labeling, aimed at reducing musculoskeletal disorders (MSDs) among office workers. Five distinct sitting postures were identified and classified with high accuracy (98.9% using KNN and 97% using Neural Networks) based on data collected from a Kinect device. The research emphasizes the importance of maintaining healthy sitting postures and offers a contactless solution for monitoring and improving workplace ergonomics.

Uploaded by

Bhaskar Rao P
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
10 views9 pages

1 - Human Sitting Postures Classification Based On Angular Features With Fuzzy-Logic Labeling

The study presents a system for classifying human sitting postures using angular features and fuzzy-logic labeling, aimed at reducing musculoskeletal disorders (MSDs) among office workers. Five distinct sitting postures were identified and classified with high accuracy (98.9% using KNN and 97% using Neural Networks) based on data collected from a Kinect device. The research emphasizes the importance of maintaining healthy sitting postures and offers a contactless solution for monitoring and improving workplace ergonomics.

Uploaded by

Bhaskar Rao P
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 9

5th International Conference on Applied

Engineering and Natural Sciences


July 10-12, 2023 : Konya, Turkey

All Sciences Proceedings © 2023 Published by All Sciences Proceedings https://www.icaens.com/


http://as-proceeding.com/

Human Sitting Postures Classification Based on Angular Features with


Fuzzy-Logic Labeling
Huseyin COSKUN*
Department of Computer Engineering, Kütahya Health Sciences University, Turkey
*
([email protected])

Abstract – Musculoskeletal disorders (MSDs) are generally associated with sitting postures. Assessing
and ensuring healthy sitting posture are indispensable aspects of reducing the occurrence of MSDs. This
study aims to develop a system that allows office workers' body postures to be contactless and recognized
by different classification methods while sitting on a chair and can be used for health applications. Five
different sitting body postures have been determined within the scope of medical and health literature
studies and relevant standards. Thirty subjects were asked to sit in these body postures for 30 seconds.
While the subjects were sitting, skeleton point position defined as a pose data of the subjects were
obtained from the Kinect device simultaneously. Five angles that are thought to distinguish sitting
positions according to different joint positions were determined and calculated. The angle values that can
take in the standard sitting position in the literature have been determined. According to these values, the
angle values in other postures were determined. A rule-based fuzzy inference system was designed using
angle values for labeling sitting posture data. Angle values were calculated to classify the labeled depth
values, and an artificial neural network classifier was designed according to these angle values. As a
result, five different sitting body postures were classified with KNN (K-Nearest Neighbours) and Neural
Network (NN), respectively, with 98.9% and 97% overall accuracy values. The study was compared with
other studies in the literature. In this context, a high-performance system design that can improve healthy
sitting behaviors for office workers that can be used in both health applications and robot vision is
presented.

Keywords – Sitting Posture, Angular Features, Fuzzy-Logic Labeling, Classification, Depth-Based Sensor, Kinect

I. INTRODUCTION number of people suffering from MSD has


Today, most people work sitting in their work increased by 25% over the past decade, accounting
environment. Sitting in the same position for a long for 2% of the global disease burden [4]. Therefore,
time or wrong sitting can be caused unhealthy body MSDs are an increasingly crucial work-related
posture, skeleton point and low back pain, health issue in contemporary workplaces.
disorders, muscle, and heart disorders, etc., leading Although, as some suggest, eliminating the
to disadvantageous situations. [1], [2] According to necessity of sitting down would be better for
the European Agency Safety and Health at Work human health [5], [6] it is obvious that it is
[3], around 60% of all workers in the EU with a inevitable to take measures such as correcting
work-related health complaint identify sitting postures and not sitting for a long time
Musculoskeletal Disorders (MSDs) as their most working. An intelligent workplace environment
serious problem. MSDs cause 58% of work-related with an automatic sitting posture tracking system is
diseases, deaths, and permanent incapacity from presented as a potential solution to save the high
work. 75% of these are the actions of sitting in the cost of health problems. In previous related studies,
wrong position for a long time. Globally, the sitting poses were determined using certain
611
skeleton point positions, and these were generally utilized eight skeleton points to recognize the
classified by machine learning methods using sitting posture of 100 subjects. While they
different features. In this context, in this study, a recognized four types of body posture, they
contactless measurement system proposal was achieved 97.77 accuracy for sitting posture. Ray et
presented for the problem of recognizing the al. [11] proposed an automated approach to classify
specified postures in line with the construction workers' postures as ergonomic or
recommendations of health institutions and experts. non-ergonomic. The dataset, which contains 22226
For this purpose, the Kinect v2 device was used. poses compiled from twelve joint body points, has
Thanks to the depth sensor and camera of the been used to classify the four postures belonging to
Kinect device, it can obtain the width, height, and 8 subjects. According to the experimental results,
depth location information of 25 joint positions. In the accuracy value of 94.8 has been achieved using
an office environment, this can provide contactless the linear discriminant analysis (LDA) method in
access to sitting posture information without real-time. The literature makes it unclear how
affecting employees. In this direction, how different sitting postures are defined or based on
different sitting poses should be classified using which medical studies or standards. Therefore, the
depth-based angular features and methods with relevant medical and health literature was
more effective performance than the studies in the examined, and the standard sitting posture was
literature were investigated in this study. defined with precise expressions and angle values
in this study. This study determined four different
II. RELATED WORKS sitting postures to distinguish the standard sitting
Paliyawan et al. [7] proposed classifying office posture from others. Besides, the difference
workers' sitting on the real-time skeleton data between the sitting posture, which the subjects feel
stream captured by a Kinect camera in an office comfortable with, and the standard sitting posture
work area. They collected 397800 poses compiled was revealed due to classification. In some studies,
from 10 body skeleton point belonging to 28 it has been mentioned that body posture classes are
different subjects to create the dataset. The determined by observation. The Kinect-based
performance of several classification methods such angular features method is proposed in this study to
as Decision Tree, (DT) Neural Network (NN), eliminate the lack of observational studies has
Naive Bayes (NB), and k-Nearest Neighbors qualitative features. In most studies in the
(KNN) have been compared. They achieved to literature, instead of classifying the data obtained
classify one postures class with 98% accuracy. with more than one method, a single method has
Thus, real-time feedback based on the three levels been proposed. Being dependent on one method
of health in ergonomics has been given to subjects. has limitations for the generalization of the study.
Pal et al. [8] researched occupational hazards from In order to eliminate this deficiency, five different
prolonged sitting in a particular employee posture. sitting poses created with angular features obtained
Sitting posture recognition has been achieved using from the Kinect device were classified with eight
seven similarity measures. Using city-block different classifiers, and the classification results
distance, they classified two sitting body posture were compared with the studies in the literature.
types with a high accuracy of 94.29% in 3.83 According to the comparison results, higher
milliseconds. Therefore, the 6500 sitting poses accuracy values were obtained by using fewer joint
containing the 16 different body skeleton points points.
from 20 subjects have been collected to create the
dataset. Bei et al. [9] present a sitting posture III. MATERIALS AND METHOD
classification method based on a Kinect device A. Determination of Sitting Postures
depth sensor. The dataset, which contains 16200
poses compiled from six body skeleton points, has The suggestions of the studies in the literature
been used to classify the nine postures belonging to [12]–[16] and definitions in ISO 7250-1:2017 [17]
18 subjects. According to the experimental results, have been referred to for the determination of
the accuracy value of 95.8 has been achieved using sitting postures. One is a healthy and standard body
the fusion of the body skeleton point features and posture defined as to suggestions in the literature
the KNN method. Li et al. [10] proposed a method and ISO 7250-1 standard. A sample drawing of a
involving BP (backpropagation) neural network. healthy and standard body posture determined
The BP network used the skeleton data captured by according to these suggestions in [12], [16], [17] is
the Kinect depth sensor to classify postures. They
612
given in Figure 1 [13]. The determined five setup was set up to create a depth sensor-based
different positions are given in Table I. sitting posture database. In this setup, the subjects
sit on a chair 1.5 meters from the Kinect device.
They were requested to exhibit the postures
defined in Table I.

Figure 1 Standard sitting illustration (a), Kinect skeleton


points (b)
In order to ensure that the sitting postures of the
subjects are correctly changed during the
experiment to apply the sitting positions specified Figure 2 (a) Data collection software, (b) experimental setup,
in Table I, a presentation containing the positions and (c) guide presentation
in Figure 4 was prepared. The subjects were asked Subjects waited for 30 seconds for each posture.
to follow their sitting postures via this presentation The real-time data collection and recognition
throughout the experiment. software (a) developed by Python, experimental
Table I. Definitions of sitting postures used in experiments setup (b), and presentation samples (c) are shown
in Figure 2. The recognition software uses the
# Position Description Ref. TensorFlow machine learning library to create
The hands were asked to sit on [12],
classification models.
Standard both armrests with the back [16],
1
sitting fully leaned back and knees [17] C. Algorithms for Calculating Joint Angles to
bent 90 degrees straight.
Label Body Posture Data
They were asked to sit, so they
Leaning to
bent forward as much as [12]– The skeleton point positions were used to exclude
2 the front
possible, avoiding contact [18] body posture transition values from the data set and
side
with the back.
It was requested that the body
to label skeleton point data according to the
be bent to the left by placing definitions in Table I. Each posture's specific
Leaning to the right foot on the left foot angle values were selected to label using the
[12]– skeleton points data.
3 the left and leaning the left arm on the
[18]
side armrest, and the contact with
the right sitting area was cut
as much as possible.
It was requested that the body
be bent to the right by placing
Leaning to the left foot on the right foot
[12]–
4 the right and resting the right arm on
[18]
side the armrest, and the contact
with the left sitting area was
cut as much as possible.
They were asked to sit and [12]– Figure 3 The skeleton points and the angles for standard
slide in the seat by creating a [18] posture
Leaning to
triangular gap in this area, in
5 the
backside
the form of cutting contact An example drawing and angle representation of
with the lower back and sitting skeleton point position data of standard sitting
back area. position is given in Figure 3. These angles are had
B. Experimental Setup and Software been defined respectively as the angle of the back
with the left upper leg axis in the sitting position
Depth sensor-based data used for this study were (A), the angle of the hip axis with the left upper leg
collected with a Kinect camera. By using the depth (B), the angle of the hip axis with the right upper
camera, monitoring can be done without the user leg (C), and the left angle of the back with the hip
needing to install any equipment. An experimental axis (D). It was decided that these angles are the
613
least number of angles that can represent incorrect 4, the changes in the regions of the joint positions
sitting postures according to suggestions in [12], corresponding to the sitting positions for different
[16], [17]. Skeleton point location data is used as axes values can be observed. The aim here is to
unit length, not actual length measures such as determine the positions of the skeleton points of
meters or inches. In order to calculate the angles A, the subjects in the sitting position and to observe
B, C, and D, four triangles given in Figure 3 were the proper sitting behavior with mathematical
formed, and the lengths of the sides forming these values instead of qualitative observation. The
triangles were calculated. Since skeleton point drawings were created as a result of axis rotation
coordinate information was obtained from the processes in order to make the skeleton point
Kinect device in 3D space, the edge lengths of the position drawings look more understandable in
triangles were calculated according to Equation 1 Figure 4. According to the angle range values
[19]. determined by the recommendations in the
(1)
literature in [12], [16], [17], the average angle
𝑑 = √(𝑥2 − 𝑥1 )2 + (𝑦2 − 𝑦1 )2 + (𝑧2 − 𝑧1 )2
values of the postures are given in Table II; it is
For a triangle whose sides are A, B, and C, seen that angle values are suitable for the postures
respectively, and each side has interior angles with in Table I.
the same name, angle A is calculated by Equation 2 Table II. Angle values of sitting postures
[19]. According to the example triangle (a) given
Posture 1 2 3 4 5
in Figure 3, whose side lengths were calculated
with Equation 1, the value of angle A was Angle A 104.3 68.8 88.2 80.4 118.2
calculated with Equation 2. Likewise, Equation 2
calculates the angles B, C, and D. Angle A 80- 80- 110-
95-105 60-90
Range 90 90 120
|𝐴𝐵|2 + |𝐴𝐶|2 − |𝐵𝐶|2 180
𝐴° =180 − 𝑎𝑟𝑐𝑐𝑜𝑠 ( )× (2)
2 × |𝐴𝐶| × |𝐴𝐵| 𝜋 Angle B 101.3 121.5 47.1 41.4 115.3

Angle B 110- 40- 40- 110-


95-105
Range 125 55 55 125

Angle C 104.2 125.7 44.2 50.5 118.2

Angle C 105- 40- 40- 110-


95-105
Range 125 55 55 125

Angle D 92.1 88.2 76.2 72.4 88.4

Angle D 70- 70-


95-105 85-100 80-90
Range 80 80

Since the sides forming the B and C angles


represent the upper legs mutually, these values
should be close to each other. When the values are
examined in Table II, it is seen that this situation
is achieved. At the same time, angle A is an angle
that should decrease when leaning forward and
increase when leaning back. Therefore, when the
values are examined, they are calculated correctly,
especially in the second and fifth sitting postures. It
Figure 4 A subject's Kinect raw data as to sitting postures is also seen that the D angle values should not
(between 1 and 2477 of supplementary data) change much since there is not much bending to
the right and left in the second and fifth sitting
During the experiment, raw data on the x postures. For each pose recorded in the dataset, the
(horizontal), y (vertical), and z (depth) axes of the angle values were calculated using the joint points
25 joint points of the subjects were obtained. The in Figure 3.
sample raw data of a subject and images of the
skeletal pattern are presented in Figure 4. In Figure
614
Figure 5 Fuzzy inference evaluation
Using SQL stored procedures, the sitting
postures were labeled according to the angle
ranges. Mamdani's fuzzy inference system is used
in the fuzzy logic-based labeling approach. Four
inputs are specified for each angle value.
According to the angle ranges in Table II, the
Gaussian membership functions and the range of
output variables for each class are defined as [0 5].
As the output variable membership function rules,
Figure 6 The labeling and classification diagram
range [0 1] for grade 1, range [1 2] for grade 2,
range [2 3] for grade 3, range [3 4] for grade 4 and 49580 sitting body posture poses were obtained for
[4 5] range is determined for the 5th grade. The 30 subjects in 5 classes. Since some subjects
output results and MSE (Mean squared error) corrected their sitting positions during the
values after the designed fuzzy inference system's experiment, the angle values obtained from the
evaluation of the angle values are presented in skeleton point data were not at the desired values.
Figure 5. When the charts are examined, it is seen Therefore, angle-based labeling could not label
that the classes are separated from each other due these data for the determined sitting position. Since
to the evaluation, and the MSE value is quite low. data obtained from different subjects were used
Therefore, the fuzzy inference result labeled angle with Kinect-based labeling, different numbers of
values clustered around their own class as data were obtained for each class. The data set has
belonging to that class. 9827, 9617, 9778, 10553, and 9805 records for
the First, second, third, fourth, and fifth postures.
D. Preparing the Dataset and Machine Learning For all classifiers, 15% of the data was used as
Methods validation and test data. The training model has
After the fuzzy-rule-based labeling process, been tested with data not previously used in
different machine learning methods classified the training. Therefore, test data is entirely different
angular features whose class information was from training data. To recognize sitting body
assigned. The labeling and classification diagram is posture, shallow machine learning algorithms, the
most widely used in the related literature, were
given in Figure 6. The depth sensor-based data of
used, and results were evaluated with performance
sitting body posture were obtained with an average
indicators. Although deep learning methods
100 ms cycle (10 Hz). The field of view is 84.1 and gradually become overwhelming, shallow
53.8 for horizontal and vertical, respectively. The classifiers remain preferred because training time is
depth distance is 3 meters. shorter than deep learning methods [20]–[22]. NN
and KNN classifiers were designed and utilized to
classify sitting posture. The classification models
with different parameters have been evaluated for
the best performance. The network has 2 inputs,
100 hidden layers, and five output layers. Data
division feature random, training function scaled
conjugate gradient (SCG), Levenberg-Marquardt
optimization method, and cross-entropy are used in
the networks. The activation function is tan-sig,
and 234 epochs were used to train. The error goal

615
has been limited to 0.001 [23], [24]. The weights
and biases are initialized using the Nguyen-
Widrow method. For the KNN, model flexibility
parameters such as the number of neighbors,
distance metric, and distance weight have been
chosen as 3, Euclidean, and Uniform, respectively.
All models were validated through a 5-fold cross-
validation (cv) process to evaluate the predictive
ability as it allows the classifier to operate without Figure 8 Performance indicators of all classifiers
bias and avoids the overfitting problem. The cv Finally, it is seen that the samples labeled as 5th
was performed without data sharing between class are mainly classified as 1st and 4th class,
training and validation data to avoid overtraining. except for their groups. When the performance of
In order to measure the performances of each the classifiers is evaluated, if Figure 7 and Figure
model, a multi-class confusion matrix which is 8 are examined, it is seen that the classification
defined in [25] and the ROC curve, is created, and accuracy of the KNN classifier is higher than the
Accuracy (A), Recall (R), precision (P), F1-score NN classifier.
(F), AUC (Area Under Curve), LogLoss (LL-
logistic loss) and Specificity (S) indicators are
calculated to evaluate performance [25].
IV. RESULTS AND DISCUSSION
For the classification of body postures,
confusion matrices for the models of NN, KNN,
and classifier methods with the highest accuracy
are presented in Figure 7. Both training and testing
processes were performed on the same computer.
Figure 9 ROC curves of classifiers
When the confusion matrices are examined, it is
seen that the samples are mainly classified
The ROC curves belonging to the models with
according to their classes.
the highest accuracy values for interpreting the
accuracy values are presented in Figure 9. When
the ROC graphs are examined, it is seen that all
models are very close to the upper left corner point
(0,1); therefore, the ability of the models to
diagnose classes fits quite well. Different studies
classify the data from different skeleton points in
the related literature. However, in these studies,
data acquired from the specified skeleton points
were classified at once, and the data from different
Figure 7 Confusion matrices of classifiers
skeleton points were not separated and classified
It is seen that the first-class labeled samples are
same time. Therefore, when True Positive Rates
mainly classified as 5th class apart from their
(TPR) in confusion matrices and Table III are
groups and vice versa. Next, it is seen that the
examined, the best and worst overall accuracy
samples labeled as 3rd class are mainly classified
classes are the second and fifth classes.
as 4th class and 2nd, except for their groups, and
the samples labeled as 4th class are mainly
classified as 3rd class, except for their groups.

616
Table III. Performances of classification methods for all this context, classification accuracy values and
posture classes other features obtained in studies in the relevant
literature were compared with the results of this
Posture

A F1 P R
Model

AUC LL S study and presented in Table IV. The feature


(%) (%) (%) (%)
extraction method involves extracting more
meaningful core data from raw data. The angular
1 0.991 0.969 0.924 0.912 0.935 0.089 0.977
calculation method is used for labeling sitting
2 0.992 0.977 0.939 0.938 0.940 0.068 0.986 poses, and angular features are used for
classification.
3 0.989 0.974 0.935 0.939 0.932 0.086 0.984
NN

Table IV Comparison of studies in the related literature


4 0.989 0.970 0.931 0.937 0.926 0.090 0.983
Ray Bei Paliyawan Li This
Study Pal [8]
5 0.987 0.963 0.909 0.912 0.905 0.103 0.978 [11] [9] [7] [10] Study

Accuracy 94.29 94.80 95.80 98.19 98.85 98.9


1 0.988 0.986 0.941 0.932 0.951 0.268 0.982

2 0.991 0.987 0.964 0.965 0.964 0.162 0.992

Time-based

Fuzzy rule
calculation

calculation

calculation

Automatic

Geometric
Labeling

calculator
Angular

Angular

Angular
Method
KNN

3 0.988 0.993 0.961 0.968 0.955 0.221 0.992

shape

set
4 0.991 0.991 0.961 0.960 0.962 0.192 0.989

Statistics features

physical features
Grayscale image
Angular feature
5 0.984 0.988 0.934 0.938 0.930 0.285 0.984

Local contour -

Human body
Feature

topological
Method

Features
Angular
In order to visualize the sitting posture data, the
data in each sitting class and sample 3D drawings
of them are given in Figure 10. When these City-
BP
Classifier Block LDA KNN NB KNN
drawings are examined, it is seen that the TP, FP, Distance
NN
and FN groups differ.
Joint
16 12 6 10 8 6
Point

The
Number 2 4 9 2 2 5
of Classes

Pose
5600 22226 16200 397800 55080 49580
Number

The
Number
20 8 18 28 100 30
of
Subjects

Dataset volume represents the total number of


postures used as training and test data. The number
Figure 10 a) TP sitting posture, b) FP sitting posture, c) FN of subjects indicates how many were collected
sitting posture while the data set was created. Joint Point is the
skeleton point value used to classify sitting
When the joint points are examined, it is seen that positions. When Table IV is examined, it is seen
most of the joint points are distinctive according to that the highest accuracy value for the models was
the sitting positions for the true positive samples. It reached in this study. While there are studies [7]
is seen that the 5, 8, and 9 joint points, which are and [10] with datasets more extensive than the
mostly fixed, are insufficient to distinguish the dataset volume in this study, other studies used
class samples from each other. The use of other smaller-volume data. As a result, high accuracy
joint points that are thought to be distinctive, such values were obtained from studies [7] and [10]
as joint no. 4, may not be included in the data set with a larger volume than the dataset volume used.
because it is difficult to obtain in terms of office The highest classification accuracy value was
workers' current positions (desk, computer, etc.). In obtained compared to the study [9], with the most
617
minor joint points in the literature. Using the same detect the skeletal points of up to 6 people creates a
labeling and classification method (KNN), a very reasonable amount in terms of system design
feature set with more joint points (6) and a data set when a single device can be used to evaluate the
with more posture class types (9) were used, and body postures of office workers. The presentation
higher accuracy was obtained compared to the of a depth sensor-based system prepares the
other study [9]. A higher validation success was infrastructure for a system that can be used in
obtained compared to the other study [10], whose intelligent robot assistants, especially in robot
data set used the same classification method (NN), vision. The system is thought to recognize sitting
with more total joint points (8) and fewer posture or other postures (bending, lifting, etc.) for those
class types (2) used. According to the sitting working in other fields. The proposed system is
posture class type, this study has more class types thought to be innovative and promising for
than half of the studies [7] [8] [10] [11] in the detecting the sitting postures of office workers and
literature. Therefore, higher accuracy values were presenting meaningful suggestions.
obtained compared to studies with the same or
lesser class types. The fact that the subjects are REFERENCES
different people means they have different body
characteristics. This may cause more separation of [1] M. S. Grace, R. E. D. Climie, and D. W. Dunstan,
"Sedentary behavior and mechanisms of cardiovascular
the classes representing the sitting posture data disease-getting to the heart of the matter," Exerc Sport Sci
obtained from each other. This situation directly Rev, vol. 45, no. 2, pp. 55–56, 2017, doi:
10.1249/JES.0000000000000107.
affects the classification success. The subjects in [2] M. S. Tremblay, R. C. Colley, T. J. Saunders, G. N.
the experiments of this study are fewer than the Healy, and N. Owen, "Physiological and health
implications of a sedentary lifestyle,"
only one studies [10] in the literature. However, it https://doi.org/10.1139/H10-079, vol. 35, no. 6, pp. 725–
is close to one study [7] and considerably higher 740, 2010, doi: 10.1139/H10-079.
than the other [11]. [3] "Health and safety statistics."
https://www.hse.gov.uk/statistics/ (accessed Mar. 12,
2023).
V. CONCLUSION AND FUTURE WORKS [4] T. Borhany, E. Shahid, W. A. Siddique, and H. Ali,
"Musculoskeletal problems in frequent computer and
Developing new, low-cost, accessible internet users," J Family Med Prim Care, vol. 7, no. 2, p.
337, 2018, doi: 10.4103/JFMPC.JFMPC_326_17.
technologies is an essential step towards [5] B. T. MacEwen, D. J. MacDonald, and J. F. Burr, "A
facilitating the assessment of sitting postures as systematic review of standing and treadmill desks in the
office workers sit for extended periods. In this workplace.," Prev Med (Baltim), vol. 70, pp. 50–58, Jan.
2015, doi: 10.1016/j.ypmed.2014.11.011.
direction, standard sitting posture has been [6] S. A. Creasy, R. J. Rogers, T. D. Byard, R. J. Kowalsky,
determined within the scope of relevant medical and J. M. Jakicic, "Energy expenditure during acute
and health studies and standards to carry out the periods of sitting, standing, and walking," J Phys Act
Health, vol. 13, no. 6, pp. 573–578, 2016.
tests of the proposed system. Depth sensor-based [7] P. Paliyawan, C. Nukoolkit, and P. Mongkolnam,
49580 sitting pose data were obtained from 30 "Prolonged sitting detection for office workers syndrome
prevention using kinect," 2014 11th International
subjects for five different sitting positions, Conference on Electrical Engineering/Electronics,
including the standard sitting posture. A fuzzy- Computer, Telecommunications and Information
logic labeling method with depth-based angular Technology, ECTI-CON 2014, 2014, doi:
10.1109/ECTICON.2014.6839785.
features for sitting position labeling data has been [8] M. Pal, S. Saha, and A. Konar, "Distance matching based
proposed for the first time in the related literature. gesture recognition for healthcare using Microsoft's
Kinect sensor," International Conference on
This method calculates the angle values between Microelectronics, Computing and Communication,
the body parts according to the sitting postures. In MicroCom 2016, Jul. 2016, doi:
order to obtain the best classification accuracy, the 10.1109/MICROCOM.2016.7522586.
[9] S. Bei, Z. Xing, L. Taocheng, and L. Qin, "Sitting posture
sitting poses dataset with the most minor joint detection using adaptively fused 3D features,"
points was classified with two different Proceedings of the 2017 IEEE 2nd Information
Technology, Networking, Electronic and Automation
classification methods. A high classification Control Conference, ITNEC 2017, vol. 2018-January, pp.
accuracy value was obtained in most of the 1073–1077, Feb. 2018, doi:
methods. In order to determine the relationship 10.1109/ITNEC.2017.8284904.
[10] B. Li, B. Bai, C. Han, H. Long, and L. Zhao, "Novel
between emotional states and sitting postures, hybrid method for human posture recognition based on
simultaneous data can be obtained by methods kinect V2," Communications in Computer and
Information Science, vol. 771, pp. 331–342, 2017, doi:
such as EEG [26], [27] and emotion detection- 10.1007/978-981-10-7299-4_27/TABLES/5.
recognition, and their similarities can be [11] S. J. Ray and J. Teizer, "Real-time construction worker
investigated. The ability of the Kinect device to posture analysis for ergonomics training," Advanced

618
Engineering Informatics, vol. 26, no. 2, pp. 439–455, Apr.
2012, doi: 10.1016/J.AEI.2012.02.011.
[12] N. J. Delleman and J. Dul, "International standards on
working postures and movements ISO 11226 and EN
1005-4," https://doi.org/10.1080/00140130701674430,
vol. 50, no. 11, pp. 1809–1819, Nov. 2007, doi:
10.1080/00140130701674430.
[13] J. Kelly, "Proper Height For Standing Desks," 2014.
https://notsitting.com/proper-height/
[14] J. Elliott, "How long should I stand at my standing desk? -
HealthPostures," 2020. https://healthpostures.com/how-
long-should-i-stand-at-my-standing-desk/ (accessed Jun.
23, 2022).
[15] Fondazione Ergo-MTM Italia, “Ergonomic Assessment
Worksheet,” 2021. [Online]. Available:
https://www.eaws.it/
[16] Canadian Centre for Occupational Health & Safety,
"Working in a Sitting Position -Good Body Position,"
2017.
[17] "ISO - ISO 7250-1:2017 - Basic human body
measurements for technological design — Part 1: Body
measurement definitions and landmarks."
https://www.iso.org/standard/65246.html (accessed Dec.
24, 2022).
[18] Y.-K. Kong, S. Lee, K.-S. Lee, and D.-M. Kim,
"Comparisons of ergonomic evaluation tools (ALLA,
RULA, REBA and OWAS) for farm work," International
journal of occupational safety and ergonomics, vol. 24,
no. 2, pp. 218–223, 2018.
[19] J. Vince, "Mathematics for Computer Graphics," 2022,
[20] J. Liu and Y. Zhang, "An Attribute-Weighted Bayes
Classifier Based on Asymmetric Correlation Coefficient,"
Intern J Pattern Recognit Artif Intell, vol. 34, no. 10, Jan.
2020, doi: 10.1142/S0218001420500251.
[21] R. K. Bania, "R-GEFS: Condorcet Rank Aggregation
with Graph Theoretic Ensemble Feature Selection
Algorithm for Classification," Intern J Pattern Recognit
Artif Intell, vol. 36, no. 9, Jun. 2022, doi:
10.1142/S021800142250032X.
[22] T. Hu et al., "Rice Variety Identification Based on the
Leaf Hyperspectral Feature via LPP-SVM," Intern J
Pattern Recognit Artif Intell, vol. 36, no. 15, p. 2350001,
Jan. 2023, doi: 10.1142/S0218001423500015.
[23] H. Coskun and T. Yigit, "Artificial Intelligence
Applications on Classification of Heart Sounds," in
Nature-Inspired Intelligent Techniques for Solving
Biomedical Engineering Problems, IGI Global, 2018, pp.
146–183.
[24] H. Coskun, O. Deperlioglu, and T. Yigit, “Ekstra Sistol
Kalp Seslerinin MFKK Öznitelikleriyle Yapay Sinir
Aǧlari Kullanilarak Siniflandirilmasi,” 2017 25th Signal
Processing and Communications Applications
Conference, SIU 2017, Jun. 2017, doi:
10.1109/SIU.2017.7960252.
[25] H. Coskun, T. Yiğit, İ. S. Üncü, M. Ersoy, and A. Topal,
"An Industrial Application Towards Classification and
Optimization of Multi-Class Tile Surface Defects Based
on Geometric and Wavelet Features," Traitement du
Signal, vol. 39, no. 6, pp. 2011–2022, Dec. 2022
[26] F. Bozkurt, H. Coskun, and H. Aydogan, "Effectiveness
of Classroom Lighting Colors Toward Students' Attention
and Meditation Extracted From Brainwaves," JOURNAL
OF EDUCATIONAL AND INSTRUCTIONAL
STUDIES, vol. 4, no. 2, pp. 6–12, 2014.
[27] H. Aydogan, F. Bozkurt, and H. Coskun, "An assessment
of brain electrical activities of students toward teacher's
specific emotions," International Journal of Psychological
and Behavioral Sciences, vol. 9, no. 6, pp. 2037–2040,
2015.

619

You might also like