Colin Lam is a software engineer at Tiliter in the Devices team. Colin has an interesting background, and even more interesting side projects. Recently, Colin worked on a research paper about new tech to track a person’s fatigue – we sat down with him to hear all about it.
“In my day-to-day at Tiliter, I work on both software and hardware in the Devices team – everything from software for inference, to hardware that ensures everything talks to each other. My background is in electrical engineering, specialising in digital signal processing, and my role at Tiliter is the perfect combination of those interests – working with CV algorithms in the IT space.”
“There’s always a challenge to be solved. We are always trying to push the boundaries on what no one has done before. There are new challenges that you don’t see anywhere else: something worth looking at, something worth trying out, something worth solving.”
“There are new challenges that you don’t see anywhere else: something worth looking at, something worth trying out, something worth solving.” Colin Lam
“Less than the big projects, I’m actually more proud of the small minor victories. It’s the little problems that are blocking progress which you can resolve and then move on that keep me going.”
“Along with my co-authors – Julien Epps and Siyuan Chen – we wrote a paper called “Wearable Fatigue Detection Using Blink-Saccade Synchronisation”. The purpose was to detect how tired a person is by eye activity. There’s been a lot of research into how blinking predicts fatigue, and also into how saccades (rapid eye movements) predict fatigue, but our paper is one of the first to blend these two areas and study their synchronisation. This can help us can develop more accurate technology to predict how tired someone is. The end goal is to develop this into wearable tech which could have significant benefits for reducing accidents on the road.
Our main finding was that blink saccade synchronisation works very effectively as a predictor and is superior to blinks or saccades alone.”
“The research we conducted doesn’t have much machine learning, but just gleans the data through a man-made algorithm. You could say it’s manual. However, with Tiliter’s computer vision we can pass information into the AI’s deep learning model. There is some similarity in the area of pre-processing – both algorithms undertake background subtraction before pushing information into the model.”
“I’m really looking forward to 2022. There’s a lot of new implementations coming alongside a large order, requiring lots of change to allow us to deploy at scale. It means a transformation of the way we do things.”
Check out our engineering opportunities here.