Lantian Li 2015-02-10
1. Verify the normalization of d-vector. That is to make pseudo speaker recognition.
The results showed that d-vector for pseudo speaker recognition has better performance than text-dependent speaker recognition.
I think this is due to the increase of training data and it is more useful than text-content information.
This foundings suggest that d-vector can be applied to pseudo speaker recognition.
2. Submit the patent application for Pro. Zheng, and wait for the acceptance notice.
1. Go on the task1.