Eye controlled wheelchair

[kotian]

hello, i want to build an eye movement controlled wheelchair using labview. could you tell me how to determine the eye position in labview? we need to check whether the eye is moving left or right or up! please help me out

[Hornless.Rhino]

One way is to use the find circular edge VI and locate the eyes and iris, then depending on how the iris relates to the eye as a whole you can determine which way the eyes are looking.

[ZcuBa]

Start out by determining where the camera would be.
If it can be head mounted, (head fixated) the issue is a matter of finding the iris location in the image, and making sure that the camera can be re-mounted at the exact same position each time the chair is used..

If not, you need face tracking, and eyetracking, so that you can track the eye relative to the face.

search the forum here for a long list of proposed solutions.

Be Aware!
eyetracked vehicles have the basic problem that the chair won't know when you want to move, and when you are just lokking at something...

so you need some gesture to start/stop the chair.

if immobilized user, this is typically a series of blinks, or similar.

[kotian]

do we need to keep the template of  the eye with it at the centre and left and right and then use pattern matchhing to determine its positon?

how do we determine whether the eye is moving left/ right using pattern matching?

[Tri_Tiep]

Hi kotian.

You can used a camera and image processing with LabVIEW and IMAQ Vision.

This topic had been a lot of paper talk about it.

In addition, You can used Electro-oculogram (EOG) to controlled wheelchair.

[kotian]

thanks for that!! 
i am suing the imaq-dx, vision acquisition and vision development modules for the implementation. i am stuck in the part where  i need to perform specific actions when the eye pattern is matched.

for example: a pattern of eye in the left side is matched with its template. i would want to give out a control signal to move the wheelchair to the left.

how do i do that?  

[BFeigum]

When your pattern match returns the coordinates of the match, you will need to determine which "quandrant" the match was found in. Once you determine that, you can use a case structure that will perform different actions depending on where the match was found in the image. In the case structure is where you would put your "control signal" output.