09-30 Lantian Li - 版本历史

2014年9月29日 (一) 02:37 Lilt

2014-09-29T02:37:11Z

2014年9月29日 (一) 02:34 Lilt

2014-09-29T02:34:32Z

2014年9月29日 (一) 02:32 Lilt

2014-09-29T02:32:05Z

Lilt：以“1. To go on studying a scoring method on GMM-UBM aiming to design a cohort reference speaker models. 1). Implement the K-means algorithem to cluster the training se...”为内容创建页面

2014-09-29T02:15:31Z

以“1. To go on studying a scoring method on GMM-UBM aiming to design a cohort reference speaker models. 1). Implement the K-means algorithem to cluster the training se...”为内容创建页面

新页面

1. To go on studying a scoring method on GMM-UBM aiming to design a cohort reference speaker models.

1). Implement the K-means algorithem to cluster the training set in order for organizing the cohort set.

2). Make the verfication score results dividied into four parts. --"Real True Speaker"/"Sensitive True

Speaker"/"Sensitive Imp Speaker"/"Abosulte Imp Speaker".

3). Re-score for the four parts on the cohort set.

4). Score ranking for each part and draw score-rank distrubution diagrams.

Next Week

1. Go on the task1 to explore the inherent law of re-scoring results and use the cohort set to

reduce the error rate on the "Sensitive True Speaker"/"Sensitive Imp Speaker".

@@ 第1行： / 第1行： @@
+Weekly Summary
 . On the basis of testing on a small scale, a total-batch experiment was made.
@@ 第18行： / 第18行： @@
 clustering algorithm is right.

@@ 第1行： / 第1行： @@
-). On the basis of testing on a small scale, a total-batch experiment was made.
 The result for "true speaker" shows that there exists a score gap between the real-speaker and cohort
@@ 第15行： / 第15行： @@
 Next Week
-. Go on the task1 to explore the inherent law of re-scoring results and use the cohort set to
+. Need to analyse the experimental results especiallly for the imposter-speaker. And to find whether the
-reduce the error rate on the "Sensitive True Speaker"/"Sensitive Imp Speaker".
+clustering algorithm is right.

@@ 第1行： / 第1行： @@
 . To go on studying a scoring method on GMM-UBM aiming to design a cohort reference speaker models.
-). Implement the K-means algorithem to cluster the training set in order for organizing the cohort set.
+). On the basis of testing on a small scale, a total-batch experiment was made.
-). Make the verfication score results dividied into four parts. --"Real True Speaker"/"Sensitive True
+The result for "true speaker" shows that there exists a score gap between the real-speaker and cohort
-Speaker"/"Sensitive Imp Speaker"/"Abosulte Imp Speaker".
+speaker. Besides, the variance from the real-true-speaker is larger than the sen-true-speaker based on the
-). Re-score for the four parts on the cohort set.
+cohort set. For imposter-speaker having the similar result, However, it does not meet the original hypothesis
-). Score ranking for each part and draw score-rank distrubution diagrams.
+that if the speaker is not the true speaker, it may have the similar scoring between this hypothesis
 Next Week