Materials Today: Proceedings xxx (xxxx) xxx

Contents lists available at ScienceDirect

Materials Today: Proceedings

journal homepage: www.elsevier.com/locate/matpr

Classification of yoga pose using machine learning techniques

J. Palanimeera ⇑, K. Ponmozhi
Department Of Computer Applications, Kalasalingam Academy of Research and Education, Krishnankoil, Tamil Nadu, India

a r t i c l e i n f o a b s t r a c t

Article history: Nowadays, yoga is that the part of existence during a number of the people. The human pose estimation
Received 16 August 2020 is that the deep rooted trouble in computer vision. That has exposed in many challenges inside the
Accepted 27 August 2020 beyond. They have many fields to capture the posture like video surveillance, biometric, webcam, sort
Available online xxxx
of the equipment, etc. The pose detection techniques have observed and it’ll be used to identify the pos-
ture and thus the accuracy of the yoga posture in machine learning techniques. To classify the yoga asana
Keywords: for Sun salutations set of postures in four machine learning models and pose estimation algorithm for a
Pose detection techniques
person’s body is used for skeleton drawing in the real-time. Sun salutations set of posture are often col-
Sun salutation set
Classify machine learning techniques
lected the important time and used for the estimation of a pose algorithm for the accuracy result of yoga
Skeleton poses. We have used for various parameters. The find the results of classify of four machine learning tech-
Angle draw nique during a sun salutation set of posture. Find the accuracy of knowledge in machine learning tech-
nique. This paper to classify the sun salutation yoga poses and which machine learning technique get
the currency results of the pose. We have collected the info in one male participant. Their age, weight,
height is often regarded. To detect the yoga pose supported the angle draw out from the skeleton joint
of the estimation of a pose algorithm.
Ó 2020 Published by Elsevier Ltd.

1. Introduction machine learning modules an outsized number of image dataset

has been created which sun salutation asana. The pose estimation
The subject is in the nature of the vulnerable and wide range algorithm has utilized in vision. This algorithm draws a skeleton
health diseases of which musculoskeletal disorders may be a cru- of a person’s body (shown in Fig. 2) by marking all the joint of a
cial arena and needs urgent attention. Once a year an outsize range body and connects all the joints which provides a stick diagram re-
of individuals are affected from various sorts of musculoskeletal ferred to as the skeleton of a body (Table 1).
disorders due to accidents or aging problem [1]. Human pose
estimation may be a challenging problem within the discipline of
computer vision. It deals with localization of human joints in a pic-
2. Related work
ture or video to make a skeletal representation. To automatically
detect a person’s pose in a picture may be a difficult task because
The review works of the physical body joint representation. The
it depends on variety of aspects like scale and determination of
foremost of works are wont to color related and captured of the
the image, illumination variation, background clutter, clothing
traditional camera. The machine learning algorithm has been used
variations, surroundings, and interaction of humans with the envi-
the model of BPNN, SVM, DT, NB classified of the techniques and
ronment [2].
used the tool of Kinect [3]. They use connect device to capture
Computer vision techniques are wont to build AI software that
the posture for coaching to the elderly people to use the trial med-
works as a trainer. This software tells about the advantages of that
icine and therapeutic service [4]. They need been wont to SVM of
pose and the accuracy of the performance. Yoga is a part of the life
the machine learning model is assessed of the physical body pos-
in many of us in recent years. This is often thanks to the advan-
ture is employed to 3D model is that the silhouette input data
tages of the health. They are doing this in right way especially in
[5]. The large of the researchers have wont to the Kinect device
right posture. Fig. 1 it gives the overview of this paper. They use
[6]. They need wont to the sensor of the image within the Kinect
device. It might to not capture the posture with the accuracy of
⇑ Corresponding author. high [7]. They even have wont to the Kinect v2. It provides to the
E-mail address: [email protected] (J. Palanimeera). high accuracy, but it require to the time is more [8].

https://doi.org/10.1016/j.matpr.2020.08.700
2214-7853/Ó 2020 Published by Elsevier Ltd.

Please cite this article as: J. Palanimeera and K. Ponmozhi, Classification of yoga pose using machine learning techniques, Materials Today: Proceedings,
https://doi.org/10.1016/j.matpr.2020.08.700
J. Palanimeera and K. Ponmozhi Materials Today: Proceedings xxx (xxxx) xxx

Fig. 1. Classification of yoga pose.

Fig. 2. Image of sun salutation asana data set.

2
J. Palanimeera and K. Ponmozhi Materials Today: Proceedings xxx (xxxx) xxx

Table 1 5. Finally get the result

Data accuracy result.

S.N Classifier of machine learning model Accuracy 5.1. Skeleton created

1 KNN 0.9902
2 SVM 0.9817 First, The image was captured and therefore the parameter
3 Naïve bayes 0.7347 value can depart 1.5. Then the new image resized to 500*500 and
4 Logistic regression 0.8461
use to the estimation of pose algorithm to urge the result for accu-
racy. The estimation of pose algorithm is often wont to the create
the skeleton of the person to perform the yoga pose. They need to
3. Human pose estimation be marked in each physical body joint. Fig. 3 shows a stick diagram.
This algorithm, it works accurate within the real time.
Human posture recognition has made huge advance-
ments within the past years. It’s evolved from 2D to 3D pose esti- 5.2. Feature extraction
mation and from single person to multi person pose estimation
[9]. Uses pose estimation to create a machine learning application Next, the pose estimation algorithm is often used. They
that helps detect shoplifters whereas [10] uses one RGB camera need physical body, it 17 joints. 12 different angle us won’t to de-
to capture 3D poses of multiple people in real-time. Human pose tect and proper the yoga poses.
estimation algorithms are often widely organized in two ways. The joint angle calculated using formula
Algorithms prototyping estimation of human poses as a geometri-
2
cal calculation are classified as generative methods while algo- a2 ¼ b þ c2  2bcðcosAÞ ð1Þ
rithms modeling human pose estimation as a picture processing
Wherea = between of point 1 and 2b = between of point 2 and
problem are classified as discriminative methods [11].
3c = between of point 1 and 3A = angle of point 2
Then find the distance of points
4. Dataset collection pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
a¼ ðx1  x2Þ2 þ ðy2  y1Þ2
The dataset is to the sun salutation yoga poses to collect to the
Where(x1, y1) co ordinate of point 1(x2, y2) co ordinate of point 2.
normal webcam.
It consists of 12 poses it include the sun salutation asana, has
5.3. Classification of machine learning technique
performed the trainer. The image is taken from the tradi-
tional room. The space is employed to the daylight area. The cap-
Used for four machine learning models, namely KNN, SVM,
ture the camera was normal webcam and to the person to
perform to the yoga within the center of the camera. The space of naïve Bayes, logistic regression. The info it is often stored. The
info need to type into training and test data set. They used for var-
the camera and the person in 4 m. Just one person’s data are
ious parameters. After the evaluation the result, it can find.
often collected. The person during the sun salutation of yoga pose
collected to real time data. The pose is selection of variables for
detection of posture algorithm it work efficiently. 6. Conclusion

5. Experiments and results This paper, to classify the sun salutation yoga poses and 4 ma-
chine learning model are classified. The yoga page is defected
The gathering of Knowledge set after the steps are followed: it supported the angle extracted from the pose estimation algo-
rithm. The accurate result’s gotten of machine learning models
1. Training and testing the image KNN. The get result of 96% of the machine learning Model. Future
2. The skeleton has been created implementation of variety of individuals (group of person) doing
3. The feature extraction many pages finding the results of SVM models.
4. Classification of machine learning
Declaration of Competing Interest

The authors declare that they have no known competing finan-

cial interests or personal relationships that could have appeared
to influence the work reported in this paper.

References

[1] M. Usama Islam, Hasan Mahmud, Faisal Bin Ashraf, Iqbal Hossain, Md. Kamrul
Hasan, Yoga posture recognition by detecting human joint points in real time
using microsoft kinect, IEEE Region Humanitarian Technology Conference R10-
HTC (2017) 1–5.
[2] L. Sigal, Human pose estimation, in: Ency. of Comput. Vision, Springer 2011.
[3] O. Patsadu, C. Nukoolkit, B. Watanapa, Human gesture recognition using kinect
camera, in: Computer Science and Software Engineering (JCSSE), 2012
International Joint Conference on, 2012, pp. 28–32.
[4] S. Obdrźalek, G. Kurillo, J. Han, T. Abresch, R. Bajcsy, et al., Realtime human
pose detection and tracking for tele-rehabilitation in virtual reality, Stud.
Health Technol. Informat. 173 (2012) 320–324.
[5] I. Cohen, H. Li, Inference of human postures by classification of 3d human body
shape, in: Analysis and Modeling of Faces and Gestures, 2003. AMFG 2003.
IEEE International Workshop on. IEEE, 2003, pp. 74–81.
[6] Xin Jin, Yuan Yao, Qiliang Jiang, Xingying Huang, Jianyi Zhang, Xiaokun Zhang,
Kejun Zhang, Virtual personal trainer via the kinect sensor, in: IEEE 16th
Fig. 3. The pose estimation can be created in skeleton. International Conference on Communication Technology (ICCT), 2015, pp. 1–6.

3
J. Palanimeera and K. Ponmozhi Materials Today: Proceedings xxx (xxxx) xxx

[7] Paula Pullen, William Seffens, Machine learning gesture analysis of yoga for [10] Dushyant Mehta, Oleksandr Sotnychenko, Franziska Mueller, Weipeng Xu,
exergame development, IET Cyber-Physical Systems: Theory Applications, 3(2) Mohamed Elgharib, Pascal Fua, Hans-Peter Seidel, Helge Rhodin, Gerard Pons-
(2018) 106–110. Moll, Christian Theobalt, XNect: real-time multi-person 3D motion capture
[8] Edwin W. Trejo, Peijiang Yuan, Recognition of Yoga poses through an with a single RGB camera, ACM Trans. Graph. 39 (4) (2020), https://doi.org/
interactive system with Kinect device, 2nd Inter- national Conference on 10.1145/3386569.3392410.
Robotics and Automation Sciences (ICRAS), 2018. [11] W. Gong, X. Zhang, J. Gonzàlez, A. Sobral, T. Bouwmans, C. Tu, H. Zahzah,
[9] P. Dar, AI guardman – a machine learning application that uses pose estimation Human pose estimation from monocular images: a comprehensive survey, in:
to detect shoplifters. [Online]. Available: https://www.analyticsvidhya.com/ Sensors, Basel, Switzerland, vol. 16, 2016.
blog/2018/06/ai-guardman-machine-learningapplication-estimates-poses-
detect-shoplifters/