Single Static Video Camera based Osteoarthritis Patients Movement Analysis Dataset

Gait recognition systems are Biometric non-invasive devices analyze the way people move. These technologies may benefit both surveillance and healthcare systems. Human Gait Recognition (HGR) is an growing area of biometric science that has provoked considerable interest in research on Co

Project Title

Single Static Video Camera based Osteoarthritis Patients Movement Analysis Dataset

Project Area of Specialization

Artificial Intelligence

Project Summary

Gait recognition systems are Biometric non-invasive devices analyze the way people move.
These technologies may benefit both surveillance and healthcare systems. Human Gait
Recognition (HGR) is an growing area of biometric science that has provoked considerable
interest in research on Computer Vision (CV). Number of factors influence gait identification
despite the specific characteristics of gait features including camera perspectives, lighting
conditions, clothing variance ,load holding, , walking speed, and shadow under feet.
Designing a framework that is strong enough to solve these obstacles is crucial for accurate
gait classification.Gait recognition algorithms can now be used in a wide range of "real-
world" applications, such as video monitoring, forensic detection and crime prevention. HGR
is used in banks, embassies and airports for security reasons .For example photos from a bank
robbery caught on CCTV were analyzed and compared to photos of the thief to identify the
robber. Furthermore, police in Sweden obtained a blurred image of the suspect from a CCTV
camera during a ban raid, which was useless for identification. Swedish authorities were able
to correctly identify the suspect after HGR examined the picture.
Model-based and model-free approaches are the two broad categories of HGR. Model-based
approaches, such as human body structure and motion models, are used to classify people.
They use structural model to characterize human subject and use basic mathematical
structure, such as intertwined pendulum, stick-figure, and ellipse fitting techniques to track
the different body parts and joint positions over time for gait. In the model-free method the
features of the object silhouette are derived from the gait period which is easy to implement
due to lower computational costs.v

Project Objectives

The objective is to create datasets that combine naturalistic recordings.
Obtaining reliable and high-accuracy data.
Improvement in physical performance.

Project Implementation Method

The methodology we are using to solve this problem is that we are creating a dataset that
combine naturalistic recording for this we take a video camera which is mounted on a tripod
then a moving object is filmed doing a motion. We will record videos in different
environment like indoor and outdoor. The recording time will be of 50 to 60 sec.By using video motion analysis software the image on screen is calibrated to the size scale
enabling measurement of real world. We do the frame preprocessing do the deep learning.
Then the feature extraction and reduction will be done like cancelling the noise, clearing the
background, detect the edges, sharpen the image, unblurr the image and many other features.
Then analysie the movement of the body.Then analysie the movement of the body.

Feature selection (FS) is an active research area in machine learning from the couple of years.
The purpose of feature selection step is to refine the high dimensional feature vector which
contains duplicate and unnecessary features for the classification. The primary goal of FS is
to eliminate redundant and unnecessary characteristics from the original feature vector to
reduce its dimensionality and pick the single most vital point for higher classification
accuracy and low computing time.
In this work, a modified whale optimizer algorithm (WOA) is used for the best features
selection. The modified whale optimization algorithm works in three phases:
Phase of Exploitation (circle technique of attack of prey/bubble-net)
phase of Exploration (searching prey)
Define new activation function for check the redundant features

25
You can take a pretrained image classification network that has already learned to extract
powerful and informative features from natural images and use it as a starting point to learn a
new task. The majority of the pretrained networks are trained on a subset of the ImageNet
database, which is used in the ImageNet Large-Scale Visual Recognition Challenge. These
networks have been trained on more than a million images and can classify images into 1000
object categories. Using a pretrained network with transfer learning is typically much faster
and easier than training a network from scratch. There are many models that can be used to
trained for this some of them are mentioned here :
MobileNetv2
Darknet-19
vgg-19
resnet-101
inseptionresnetv

nasnetlarge
vgg-16

Benefits of the Project

The motivation of working on this is that so we can improve patient care Over the last few
years, a lot of research has gone into defining a person by their gait. In the gait recognition
method, deep learning-based techniques, silhouette extraction-based gait recognition, and
classical features-based techniques were all used. As we know osteoarthritis is a disease that
effect your knee and once its detected we can diagnose issue causing pain. We Analyze
osteoarthritis problem and identify individuals unique movement. There will be availability
of dataset so it will be easy for other people to work on the subject. The gait patterns can be
determined. Deep learning-based approaches for high-level descriptors have been described
in the literature. Poor lighting, a single person carrying things, clothing, and a variety of view
angles are all common issues in the gait recognition process.

Technical Details of Final Deliverable

Dataset: This database includes 4 different angles, including 90, 120, 270 and 360 degrees,
with three different styles of walking sequences including a regular walking sequence,
walking with a backpack and walking with pain. Each video of the gait sequence is captured
at a rate of 25 frames per second with a resolution of 352 240 pixels.

Final Deliverable of the Project

Software System

Core Industry

Medical

Other Industries

Health

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