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 add new speaker for enrollment? #1

Closed Sadam1195 closed 4 years ago

Sadam1195 commented 4 years ago

What modifications do I need to do in order to add new speaker using enroll.py file. @jymsuper

jymsuper commented 4 years ago

In 'split_enroll_and_test' function, we get a list of both enroll and test files. It is denoted as 'DB_all' and it is a pandas dataframe. According to the filename, we split the list and get lists of enroll and test files, which are 'enroll_DB' and 'test_DB' respectively.

In 'feat_logfbank_nfilt40/test', I put both enroll and test files for each speaker. In each speaker folder (e.g., 103F3021), you can see both 'enroll.p' and 'test.p'. When you need to add a new speaker, create a new speaker folder and add 'enroll.p' file in the folder.

Sadam1195 commented 4 years ago

Thanks for clearing that up.