Methodology

Tools Used:

Microsoft Kinect, Visual Studio, C#, XAML, Windows SDK 2.0

The Kinect is a motion capture device that can recognize 25 different joints along the human body. We decided that the best way to track and compare dance moves is to do it by differences in angles as opposed to length or position. This way, we do not have to worry about different body proportions, heights, etc.

We used 17 joints to calculate 11 angles. Most of the joints we ignored were at the center of the body and the fingertips.

marinMachine.mp4

In our UI, we placed an actual image of the move next to the responsive skeleton. Since we do not take into consideration the z-axis, the image helps the user get a general sense of the move that they have to match.

User Feedback

Each joint (green circles) that we calculated would vary in size depending on how close that specific angle was to the original. Large circles meant that the angle was far away from the original. Small circles meant that the angle was close to the original. This feedback would let the user adjust their position until it matched.

Problems/Issues

Despite how impressive the technology is, the Kinect has a number of innate flaws with body recognition.

When limbs are too close together, or too close to the body, the Kinect loses a lot of accuracy.
- aka. overlapping body parts can cause issues
Additionally, while the Kinect can supposedly recognize ankles and has 3 hand joints, it struggles to actually recognize them consistently
While some joints lacked consistency, others were static and didn't change their angles at all, including 3 centerline joints.

Report abuse