YOGGUIDE: DETECTING, RECOGNIZING AND TRACKING YOGA POSES USING VISION TECHNIQUES

Analyzing and understanding human poses estimation is a subject that has been studied extensively in the past two decades. Therefore, this represents interest for many promising applications in different domains, such as security, video games, physical therapy, etc. Problems arise when monitori

Project Title

YOGGUIDE: DETECTING, RECOGNIZING AND TRACKING YOGA POSES USING VISION TECHNIQUES

Project Area of Specialization

Artificial Intelligence

Project Summary

Analyzing and understanding human poses estimation is a subject that has been studied extensively in the past two decades. Therefore, this represents interest for many promising applications in different domains, such as security, video games, physical therapy, etc. Problems arise when monitoring human movement is required. One of the challenges in post estimation is yoga. Nowadays, Yoga has become a well-known discipline around the world which is a safe and effective exercise to increase physical activity, especially, in strength, flexibility, and balance, to boost physical and mental well-being.

As we all know yoga is said as the art of relaxation. From last year, people are dealing with greater pressure and suffering from stress faster than ever before. Today international problem is not hunger, poverty, drugs or fear of war. It is tension, hypertension. One who knows how to free from tension and how to control it he will be able to control high blood pressure, heart disease etc. Such stresses and tension accumulate in the various layers of human personality. We concentrate in the physical, intellectual, and emotional systems. Yoga solves tension issues with a big periscope.

Now problem rise when people start doing yoga at home without an instructor because they can’t afford, or join a yoga class due to transportation and timing schedule issues. Yes, Yoga has many benefits but doing wrong posture can cause serious harm to health including backache, ankle sprain and muscle pulls etc. So, for that we are making a desktop application which will alert you if your posture is incorrect. In this project we are going to cover 3 main yoga poses namely Goddess Squat, Warrior and Reverse Warrior. We will cover yoga poses according to body type i.e. slim, fat and size with Kinect device. At last, the output of this project will be a system through which user can do yoga exercises without any instructor with correct posture recommendation.

Project Objectives

Our project objective is:

• To provide an application for correct yoga poses
• To work on correct pose estimation using real time image capturing

Project Implementation Method

Our methodology is divided into three parts; data acquisition, data processing, yoga posture detection.

• Data Acquisition:

Data is gathered through volunteers by the students where volunteers would perform three yoga poses defined in the scope. Data is than trained on around 100 pictures which will be implemented on real time capturing frame by frame. This is where we need Microsoft 3D sensor Kinect. For capturing video data, Kinect is well equipped with its in-built infrared laser projector that is combined with a monochrome CMOS sensor. This device also has a multiarray microphone and an RGB camera that facilitates to capture image. Apart from ability of Kinect to facilitate color image and depth image, it also has a skeletal tracking tool that can recognize 20 joints of a human body. Each joint point saves three information; index value, coordinates along x, y and z direction and status of joint. The index value is unique for each joint. The co-ordinates provide the value in x, y and z direction of the joint. The status information provides whether the joint is tracked or not.

• Data Processing:

We are using 15 joint points for data processing and not all 20 points because our data is only based on 15 joints. The joint points will be given in the pictorial form in the report. And since we are working on three yoga poses we will predefine and mark the joint points.

• Yoga Posture Detection:

For yoga posture detection, we will mark 10 angle points which we will be used for calculations. All the x, y and z coordinates of the joint points determines the structure of every single yoga position. We will take a standard reference model and compare it with our candidate frames. If our model deviates too much from the reference model (given threshold), recognition will be unsuccessful. If not, pose will be successfully recognized.

Deliverables:

A desktop application is to be delivered as the final outcome which recognizes correct yoga poses in real time using Kinect sensors.

Flow Chart:

Benefits of the Project

This project we will make a system to recognize three major yoga pose by detecting of human joint points using Kinect device. We have detected the yoga poses considering above 97% accuracy in every angle between different body parts. Our system can also be used to recognize other Yoga poses from the reference model of each pose. Therefore, we hope our proposed approach will help to practice yoga without trainer.

Technical Details of Final Deliverable

A desktop application is to be delivered as the final outcome which recognizes correct yoga poses in real time. Its a hardware and software integrated as we there is need of camera for our application.

We will use kinect device for our dataset to detect joint points

Final Deliverable of the Project

HW/SW integrated system

Core Industry

Health

Other Industries

Core Technology

Artificial Intelligence(AI)

Other Technologies

Others

Sustainable Development Goals

Good Health and Well-Being for People

Required Resources

If you need this project, please contact me on contact@adikhanofficial.com