search

microsoft / UniSpeech

UniSpeech - Large Scale Self-Supervised Learning for Speech
Other
406 stars 71 forks source link

Getting speaker embeddings #13

Closed AH289 closed 2 years ago

AH289 commented 2 years ago

UniSpeech-SAT directory in this repo contains an example. The example takes a .wav file as an input and produces a tensor 'f' as an output. Can I get the speaker embeddings from 'f'?

MarkWuNLP commented 2 years ago

UniSpeech-SAT directory in this repo contains an example. The example takes a .wav file as an input and produces a tensor 'f' as an output. Can I get the speaker embeddings from 'f'?

Hi, If you would like to get speaker embeddings, you can refer to what we have done for speaker verification https://github.com/microsoft/UniSpeech/tree/main/UniSpeech-SAT/speaker_verification . @czy97 Add Zhengyang into the thread.

czy97 commented 2 years ago

UniSpeech-SAT directory in this repo contains an example. The example takes a .wav file as an input and produces a tensor 'f' as an output. Can I get the speaker embeddings from 'f'?

Specifically, you can get the speaker embedding from this line: https://github.com/microsoft/UniSpeech/blob/866ad28bdb3615263d4f77d36a64fbe564a412b0/UniSpeech-SAT/speaker_verification/verification.py#L50

AH289 commented 2 years ago

Thank you so much.