search

jymsuper / SpeakerRecognition_tutorial

Simple d-vector based Speaker Recognition (verification and identification) using Pytorch
MIT License
210 stars 46 forks source link

how to calculate EER in this code? #9

Closed ASR2020Guru closed 3 years ago

ASR2020Guru commented 3 years ago

Hi @jymsuper ,

Thanks for sharing this excellent codes.

I have go through the identification.py and verification.py files for calculate the perfermanes after the enroll process. Could you give me some ideas about how to calculate the EER ?

Many thanks

jymsuper commented 3 years ago

Hi. Thank you for your interest in my work. I am sorry, but I do not upload the code about calculating the EER.

(1) To calculate the EER, you need to make a trial file for verification. You can see the example below. https://www.robots.ox.ac.uk/~vgg/data/voxceleb/meta/veri_test.txt

(2) In each line, there are two wav files. You have to extract the speaker embedding vector for each file. Then, compute the cosine similarity between the embedding vectors.

(3) Using the code below (from kaldi toolkit), you can calculate the EER. cf )scores : cosine similarity, labels : label from the trial file.

https://github.com/kaldi-asr/kaldi/blob/master/egs/sre08/v1/sid/compute_min_dcf.py

ASR2020Guru commented 3 years ago

great, thanks for your reply @jymsuper

It is very helpful👍