Action Recognition Based on Depth Motion Map and Hybrid Classifier

Algorithm 1

The proposed ELM-CRC classifier for action recognition.

Input: DMMs-based disCLBP features of the test image y; DMMs-based disCLBP

features of the training action database with 20 classes;

Output: The class label of .

(1) Training the ELM classifier using the action database ;

(2) Calculate the network output by ELM classifier;

(3) Find the first and second largest entries of ELM output and ;

(4) if then

(5)  

(6) else

(7)  Set the indexes of k largest entries in ;

(8)  Get the sub-dictionary = [] in the action database ;

(9)  Solve

(10) end if

(11) for to do

(12)  Calculate the residuals ;

(13) end for

(14)