Difference between revisions of "140707 - Lantian Li"

From cslt Wiki
Jump to: navigation, search
(以内容“Weekly Summary 1.Two groups of experiments were made under the condition of different window_size and shift_size. And found that after unified shift_size and fixed th...”创建新页面)
 
Line 1: Line 1:
 
Weekly Summary
 
Weekly Summary
  
1.Two groups of experiments were made under the condition of different window_size and shift_size.
+
1.Make a new database to train the UBM and re-implement the experiment from last week.  
  
And found that after unified shift_size and fixed the number of frames,female is more sensitive than male.
+
Found that although we had modified the UBM training tools,accoring to the HTK,which could avoid the code-empty problem,  
  
Male presents a certain stability, while female in 20-10ms to 24-12ms exites an obvious turning point and has strong sensitivity to window_size and
+
it still had effect on the system performance.
  
shift_size.
+
2.Using the re-weighted GMM-UBM system to make digital speaker recognition.
 
+
2.Using the re-weighted GMM-UBM system to  
+
  
 
Next Week
 
Next Week
  
 
1.Go on the task1 and task2.
 
1.Go on the task1 and task2.

Revision as of 15:00, 7 July 2014

Weekly Summary

1.Make a new database to train the UBM and re-implement the experiment from last week.

Found that although we had modified the UBM training tools,accoring to the HTK,which could avoid the code-empty problem,

it still had effect on the system performance.

2.Using the re-weighted GMM-UBM system to make digital speaker recognition.

Next Week

1.Go on the task1 and task2.