funcwj/ge2e-speaker-verification

Pytorch implement of "Generalized End-to-End Loss for Speaker Verification"

1. Vad (recommend py-webrtcvad)
2. Log mel-spectrogram features (recommend librosa)
3. Prepare data as data/{train,dev}/{feats.scp,spk2utt}

see train.sh and compute_dvector.py

Wan L, Wang Q, Papir A, et al. Generalized end-to-end loss for speaker verification[C]//2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2018: 4879-4883.