Weekly Update 4/28/2011

What did we do this week:

There was not much more done this week on the primary Kinect app due to time constraints.  Small bug fixes were completed but nothing too severe.

The third party app, however, made considerable progress.  There were some enhancements made and some major bugs were fixed.  At this stage it is pretty close to being complete.

Problems:

No significant problems other than time.

Project Status: 

The project is starting to wrap up and is pretty close to completion.  There are only a few more things left to do.  We want to integrate at least one gesture into the third party app but we are a little concerned that we wont have enough time to complete this.

New Ideas:

No new ideas as the project is pretty much done.

4/21/2010

What did we do this week:

This week we finally nailed down the interactions and gestures that we are using for the Kinect system.  The mouse is tracked with the right hand while the left hand does the clicking and other gestures:

left hand straight: click

left hand up: right click

left hand down: double click

left hand right: alt-tab

These gestures were run through a user study with 10 users and the report was written up.  The new system was much better received than the old system so that was encouraging.  The picture below is the template that we had users click on in MS Paint for the user study.

Problems:

The lack of precision from the Kinect is still a problem.  The hand tracking is pretty jittery but unfortunately that is just not something that we can fix in the current time frame.

Project Status: 

The project is entering its final phase.  We are on schedule to be able to deliver a system that will allow users to control the windows environment.  Third party application support is questionable.

New Ideas:

No new ideas.  It's a little too late for that.

Weekly Update 4/14/11

What did we do this week:

This week involved a good amount of more refactoring.  We worked some more on the clicking mechanism and changed it twice in the week.  Originally we had planned on making the clicking with the right hand work better but that just didn't seem to work well enough.  At the end of the week we got multi hand tracking working so now we can click by pushing in with the left hand while the right hand is used only for tracking.

The client app also made some considerable progress.  It can now handle pictures, videos, and audio.

Problems:

We are still having trouble getting the clicking down just right.  This has really put us on hold because fixing this has to be our first priority.

Project Status: 

The project is coming along but it's not shaping up into what we originally had in mind.  We had initially envisioned a system that was more geared towards third party developers but we now realize that we must focus more on the end user.

New Ideas:

This week we had the idea of using the left hand to click while the right hand was tracking.

Video:

File format is giving YouTube trouble.  Video can be downloaded at this link:

http://dl.dropbox.com/u/280520/ClientAppVideo_4-13.avi

Weekly Update 4/7/2011

What did we do this week:

This week has been spent trying to improve the clicking mechanism.  Last week we learned in our user study that the clicking wasn't necessarily done correctly.  People had a good amount of trouble performing accurate clicks.  To address this we are now locking the X and Y coordinates when we see a fast change in Z.

Problems:

When running the user study we found that our clicking mechanism was pretty bad to say the least.  The problem is now to fix it.  Due to this bug we are having to change the scope of our system.  Too much of our system was dependent on having a working click and track.

Project Status: 

The project scope has been greatly reduced in the past week.  Due to the accuracy problems and the time needed to fix this, we have decided to drop third party application support for now.  We recognize that no matter how well we support third party apps, no one will want to use our system if we don't have a good system mouse replacement.

New Ideas:

Currently we are locking the X and Y coordinates when a swift change in the Z coordinate was detected.  While this provided a large increase in clicking accuracy, it is still not good enough.  We are now thinking that we will also do some backtracking on the points that the hand was at.  Therefore, when the user starts a click,  the program will look back 5 or so points and use that point as the point that the user wanted to click.