Cycle 2 – Hand tracking algo

I spent the majority of the time leading up to this cycle devising a hand-tracking algorithm that could efficiently and robustly track the position of the fingers in (one) hand. There are a few approaches I took:

Approach 1: Get the position of the palm of the hand from polling the leap motion controller. Then subtract the difference of the finger position to get its relative position.

Pros: Consistent, deals with location away from Leap motion controller well

Cons: Gives bad data at certain orientations

Approach 2: Get the differences of the finger positions in relation to each other. These deltas will provide relative finger positions to each other.

Pros: Works well at most hand orientations

Cons: Not as good when fingers close together

Approach 3: Check to see if each finger is extended and generate a unique code corresponding to each finger for each frame.This is the approach I went with. This makes it much easier to program as it abstracts away a lot of the calculations involved. Also, this ended up being arguably the most robust out of the three due to only using Leap Motion’s api.

Pros: Very robust, easy to program, works well at most orientations

Cons: Will have to make hand signals simpler to effectively use this approach.



Leave a Reply