Research Article

Developments of Machine Learning Schemes for Dynamic Time-Wrapping-Based Speech Recognition

Pseudocode 2

The pseudocode of priority-rejection learning.
Procedure DTW_Priority_Rejection_Learning ();
/*DTW recognition process*/
Perform DTW template matching and store DTW distance;
Output recognition result before learning;
/* A decision of DTW system learning made by a supervisor */
If (Decision == “YES”) then
/* Correct recognition and then start the learning process */
Label the recognition result and set the index ;
/* is a relative index in referenced templates database */
If ( setting == TRUE) then
For each
 /* is numbers of the estimated DTW-distance with */
  Search the worst (the largest) DTW-distance;
End For
 Return the template item with the worst DTW-distance;
 /* Removal process */
 Delete the found template item with the worst DTW- distance in the database;
 /* Process of learning data */
 Convert the testing data to be the learning data;
 Feature extraction;
If (Removal process finished == TRUE) then
  Add into the referenced pattern index;
  For each frame ( to total frames of the learning data)
   Add featurest into reference templates;
  End For
End If
End If
/* End of learning process */
Else
/* End of learning process (No learning) */
End If