Let's assume that you are trying to implement the simplest form of a piano, and only need to produce a single note at a time.
Let's also assume that you are not going to try and make this an 88 key piano, with white and black keys. Let say our piano will be one octave, eight keys, C D E F G A B C.
Assuming you already have a way to read values from the Kinect, these are going to most likely be in millimeters, three axis. So (X,Y,Z), where X measures left to right, Y measures front to back, and Z measures up and down.
You only need to move your hand left to right, and see the range of X values that the device returns. This will be your X-Range. Lets say that on the left, you get a value of -500, and on the right you get a value of 480. Your X-Range is 480 - (-500) = 980. Divide 980 by 8 to get your key spacing 122.5. 122.5 now becomes your key multiplier
If X <= -500 + (122.5 * 1), then C
If X <= -500 + (122.5 * 2), then D
If X <= -500 + (122.5 * 3), then E
If X <= -500 + (122.5 * 4), then F
If X <= -500 + (122.5 * 5), then G
If X <= -500 + (122.5 * 6), then A
If X <= -500 + (122.5 * 7), then B
If X <= -500 + (122.5 * 8), then C
Now you just need to look at Z values, to determine when the hand has touched the table. Since you probably want to give a little wiggle room, set the Key Touch Z heigh a little bit higher than the highest sensor value you get when you move your hand On the table left to right.
If Z <= -10 and X <= -500 + (122.5 * 1), then C
If Z <= -10 and X <= -500 + (122.5 * 2), then D
If Z <= -10 and X <= -500 + (122.5 * 3), then E
If Z <= -10 and X <= -500 + (122.5 * 4), then F
If Z <= -10 and X <= -500 + (122.5 * 5), then G
If Z <= -10 and X <= -500 + (122.5 * 6), then A
If Z <= -10 and X <= -500 + (122.5 * 7), then B
If Z <= -10 and X <= -500 + (122.5 * 8), then C
