14-10-20 Lantian Li
1. Statistical classification results, comparion with the eer result.
The results show that for the FA, using SVM is better than the baseline by eer, but FR is the opposite.
At first, I think the results related to the SVM training data. However, after adjusting the selection criteria,
the result is still the case.
2. Comparion the SVM results with the eer of t-norm, the results show that the number of FA + PR is similar.
3. Read the book "speech communication", understand the speech prodution and acoustic phonetics.
1. With the additional knowledge from cohort, increasing the feature dimension by adding cohort score and score difference.
2. Increase the number of SVM iterations.
3. Go on reading the book "speech communication"