Difference between revisions of "Lantian Li 15-04-20"

From cslt Wiki
Jump to: navigation, search
(以“Weekly Summary 1. Prepare to construct a cohort-based SVM classifier on score-level. 2. Using the "Elbow method" to determine the number of clusters under K-means...”为内容创建页面)
 
Line 1: Line 1:
 
Weekly Summary
 
Weekly Summary
  
1. Prepare to construct a cohort-based SVM classifier on score-level.
+
1. Search for discriminative features for the cohort-based SVM classifier.
  
2. Using the "Elbow method" to determine the number of clusters under K-means algorithm.
+
We select orignal score, t-norm score, relative ranking, cohort-score, cohort-subtraction-score(delta-score).
  
Finally, we choose 10 as the number of K-means clusters.
+
2. Verify the discriminative ability of the relative ranking and delta-score.
  
3. Help Pro.Zheng collect dialect speech.
+
Results show that these two features can be applied for SVM classification.
 +
 
 +
The best performance is obtained when SVM input feature is made up of orignal score, t-norm score, relative ranking and delta-score.
 +
 
 +
3. Help Pro.Zheng collect digit speech.
  
 
Next Week
 
Next Week
  
1. Go on the task 1.
+
1. Go on the task 1 and 2.

Revision as of 14:34, 20 April 2015

Weekly Summary

1. Search for discriminative features for the cohort-based SVM classifier.

We select orignal score, t-norm score, relative ranking, cohort-score, cohort-subtraction-score(delta-score).

2. Verify the discriminative ability of the relative ranking and delta-score.

Results show that these two features can be applied for SVM classification.

The best performance is obtained when SVM input feature is made up of orignal score, t-norm score, relative ranking and delta-score.

3. Help Pro.Zheng collect digit speech.

Next Week

1. Go on the task 1 and 2.