Ahmed Ahsan Khan, Abeer Alsadoon, Shatha Habeeb Jafer, P.W.C. Prasad , Oday D. Jerew, Paul Manoranjan, “A Novel Hybrid Fall Detection Technique Using Body Part Tracking and Acceleration”, IEEE International Conference on Innovative Technologies in Intelligent Systems & Industrial Application (CITISIA), Sydney, 2020. Recently Accepted
It took me 3 years to finally receive this good news through an email from my Professor Dr Abeer Alsadoon. It was a proud moment for me as it was one of my goal since I started writing my thesis report for my Master’s degree. It took few rejections over the years before it could finally be accepted. Being the first author of the paper makes it ten times better. The paper will be published in January 2021.
When I started my blog here on medium I mentioned how Java could be used to implement Machine Learning / Artificial Intelligence. That article gained a lot of views, mostly from Google Search (87%), and it is by far my most viewed article on this platform.
I used Java for Machine Learning in my thesis as mentioned in my article above. It was a challenge for me and a brave move I must say because Java is not a popular language for ML and I had to face challenges along the way. Now that when my same thesis report is selected in my IEEE Conference I can confidently say that I implemented something which is not a common practise and I succeeded in doing so.
You might have heard or seen people say that to be successful in something you should be hardworking, or in some cases they say that smart people outplay hard-workers because they know how to do their things smartly and also because they have more knowledge. I think that on top of both of these things, attitude towards work is what makes you successful.
When I started writing my thesis paper, I showed interest to my professor that I am interested in publishing my work in one of the IEEE journals or conference and I worked hard to get High Distinction grade in my thesis course. My attitude towards my work and the research which I did, made Dr Abeer so considerate that for last three years she has been trying to publish my article which finally got accepted. If I sum up how this all happened I would say that there were people more knowledgeable than me and more hardworking but attitude towards my work is what finally gave me what I had hoped for. As Dr Abeer mentioned in her email to me that she is more happier than me because she knew what it meant to me to get this work published and for the same reason she didn’t lose hope and fought for 3 years despite few rejections along the way.
My research is titled “A Novel Hybrid Fall Detection Technique Using Body Part Tracking and Acceleration”. During my work I had researched 25 articles from different journals on the same topic to get insights of their work and what was actually tested. Then out of those articles I had to choose two best solutions and find a solution of my own. I found an online dataset which had data for part tracking and acceleration and therefore I used those data to implement my algorithm in Java. Specifically I used SVM (Support Vector Machine) from Java-ML library to implement my algorithm.
The best part of my thesis was that there was no limitation of what language I have to use to proof my work. This gave me independence to choose Java as my language of choice which few would have done. In my thesis I proposed a hypothesis that stated “A fall event exists when there is an abrupt change in acceleration followed by a decrease in the vertical distance between head and body centre.” This combination helped us to detect falls perfectly, and we could distinguish them from other activities of daily living such as bending down or lying on a floor. The algorithm I created worked better on the dataset on which I tested compared to the current best solution that was selected from my literature review.
In the end, we have to understand that realistically speaking there is no such thing as perfect solution. Things evolve and there is always room for improvement with new technology and algorithms being invented and created on almost daily basis. As of now, with the hypothesis presented, the solution mentioned in my journal article is considered as the best.