Mining Key Skeleton Poses with Latent SVM for Action Recognition

Algorithm 1

Require:

 ;

 ;

 for do

  where is the pos-th frame of the first video in

  for do

   where pos –

  end for

  Train with pos_pose and neg_pose

  for do

   For each frame ,

  end for

  pos =

 end for

 Training and with linear SVM