acrandal
commented
2 months ago I completely missed this issue submission and now I find it a whole year later. I'm sorry that I didn't catch this sooner.
I can see what you're attempting to do, though I don't think this sensor has the resolution to accomplish the described goal. Mice and trackballs use lasers with a very high and accurate feedback of change of distance/surface colors to detect changes (the ball rolling or surface moving in front of the laser).
This sensor is a two dimensional array of infrared sensors. Their feedback is relatively slow and low resolution in the face of changes. The output of the device that I have been able to find so far is the hardware's output of using the embedded centroid clustering to determine the mass center of an occluding object (your finger/hand, etc).
If you could get the actual sensor data values from the 30x30 IR sensor array, you could get closer to an image of the object in view. That's my issue about Image Mode, which the documentation mentions, but I haven't been able to figure out how to get all 900 data points via I2C. Even if you had those values, it would unlikely give you the resolution to determine a rolling object in view unless that object had large protrusions or significant head surface differences. Even then, it's not a very fast sensor (IR generally isn't).
I would look more closely how mice track moving surfaces and even the possibility of pulling apart a laser mouse and building a frame to hold the laser in front of your rolling surface while watching the output of the mouse. There's not much hardware in a mouse, so it might fit into a reasonably small space.
SinanAkkoyun
Hello, I understand that this might get out of topic pretty fast but I am trying to understand if a gesture recognition sensor and library like these can recognize continuos ball rotations in any direction like you would expect, for example, from a trackball.
To be clearer: would the gesture recognition chip in cursor mode behave like an optical navigation chip for a mouse? The sensor documentation is not that extensive in explaining this specific mode.
To my understanding, in cursor mode, the chip simply considers the space it sees like (at least) a bidimensional field where a moving object moving across the visibile space boundaries is reported as being positioned at specific X/Y coordinates. If this is the case would it be possible to trick it into a different behavior?