LIS is a medical condition where a person is almost completely paralysed. People with LIS cannot move any part of their body besides their eyes. The condition can be caused by a variety of diseases including ALS, Heart and Stroke, Multiple Sclerosis. LIS victims have great difficulty moving in their environment and require constant assistance.
Just because people have difficulty moving through physical space, should not mean they also have difficulty travelling through cyberspace.
FreedomBlink is a software tool that allows LIS patients to surf the web. The FreedomBlink application is operated entirely by eye blinks, ensuring that LIS patients can use the software with complete independence.
The United Nations declared that internet access is a human right. We wanted to ensure that the internet was accessible to all members of society.
How we built it
OpenCV was used for the face and eye detections, specifically Haar Cascades. NLP and machine learning was used to provide a predictive text feature to the website, specifically the PyAutocomplete library. The GUI was designed using PyQT4.
Challenges we ran into
Too Many, can’t list them all :) It was extremely difficult designing an accurate eye detection tool. We often had issues with inconsistent lighting affecting the accuracy of the detectors. We made our eye detector more accurate by adding a hough transform function to the detector.
We were not sure how the text interface for our application should look like. We discussed several different ways to blink text from morse code to binary select.After researching existing Eye-To-Text solutions, we settled on a grid-based system. We found that this was the most efficient way to blink text.
Accomplishments that we're proud of
Currently, there are existing commercial products that allow people to navigate the web with eye movement. These products rely on expensive hardware that cost tens of thousands of dollars. We wanted our solution to not only be accessible for people with LIS, but for people of all financial classes as well.
Our software can be run on any low end laptop. The only required peripheral is a basic webcam. We’re proud that financial constraints do not prevent people from using our software.
What we learned
Our software uses a wide variety of computer vision concepts, including hough transform, guassian blur, and haar cascades.
What's next for FreedomBlink
Currently, FreedomBlink does not contain all of the intricacies and features of a modern web browser. We would like to expand our interface to include these features, such as copy/paste, back button, and writing emails. Additionally, we would like to expand our blink detector software to detect eye winks as well.