Lantian Li 2015-12-07
1. Go on my deep speaker embedding tasks:
1). DNN-ivector -- Done !
As shown in CVSS 496.
It can be seen that the DNN-ivector framework has promising results even using Chinese ASR decoder.
2). SUSR paper.
The parallel scoring scripts have alread done and still preparing for the unit-based experiments.
As shown in CVSS 425.
3). AutoEncoder + metric learning. -- Hold
1. Go on the task 2) and 3).