Surf and Adaboost for hand recognition

Hello,

I´m trying to implement an algorithm of this paper:

http://www.bmva.org/bmvc/2012/WS/paper5.pdf

It´s called "Training process for all target postures" algorithm, in the page 6.

Basicly it´s a technique that use adaboost and SURF (extractor of features) for hand posture recognition. Like I wrote before, I´m trying implement but i still didn´t understand this algorithm. The problem are some question that i have, like:

1. What is the matching score list, mention in page 5, last paragraph?
2. What is the relationship between threshold and the training process (trainning algorithm)?
3. In the line 7 of training algorithm: Which is the return of the function ht(Ix,ft,tet)?

I would like to know if someone have this algorithm implemented or if someone can help me in this questions.

Thank you very much for the answer (help), because i´m already investing some time in this algorithm and i still didn´t understand this algorithm to implement.

I inserted the link of the article in the post but it´s broken. The real link is:

www.bmva.org/bmvc/2012/WS/paper5.pdf

Sorry about this problem.