Here’s an interesting hack from France – a gesture recognition system that is currently at the starting stages of what could become a sign language translator!
The hack uses OpenNI and NITE Libraries in conjunction with a Kinect to capture certain gestures which is linked to a Fast Artificial Neural Network (FANN) library that serves as the recognition system. Based on the video, the capture system doesn’t seem to recognize hand and finger positions at the moment (which are important factors in sign language); currently, the application can only recognize and translate “hello” and “sorry”.
The team believes that the “coding part is now over” and are confident that they can teach the neural network more words and gestures, allowing them to eventually turn the hack into a practical translation tool. The developers also foresee other uses for the application such as controlling computers through gestures. The video below shows the hack in action. While it is in French it does have English annotations so you won’t have any issues following it.
This is an amazing hack that, once perfected, can have a lot of real-world applications. This is one tool that we’ll definitely be keeping an eye on.
While we certainly appreciate the idea and effort, we have to point out that you actually signed “thank you” instead of “hello.”
And “sorry” is signed on chest level, not on the stomach.
If you end writing a program we hope that you will get help from a certified interpreter to avoid mistakes like these.
I hate to inform you, “Deaf Service Bureau,” however, in French Sign Language, Bonjour is exactly as they signed it in this video, as well as desolé. Don’t forget that there are many types of sign language in the world, not just ASL.
using Omek’s gesture authoring tool i believe your work would be 10 times easier and faster, whatever the gesture would be.
http://www.omekinteractive.com/
plus, i agree that you should make a research before starting to work on that kind of project.
Good luck and great idea for making it easier computer access for the deaf.