SRE16 Xvector Model
An xvector DNN trained on augmented Switchboard, Mixer 6, and NIST SREs. The directory also contains a PLDA backend for scoring.
SRE16 Xvector Model 1a
- Date
- 2017-10-04
- Uploader
- David Snyder
- Recipe
- egs/sre16/v2
- Kaldi Version
- e082c17
- Model Type
- Speaker recognition, Xvector TDNN
- Error Rate
- 8.57% EER
- Notes
- Evaluated on the pooled SRE16 test set. Notes on the model and training data used
