search

Janghyun1230 / Speaker_Verification

Tensorflow implementation of "Generalized End-to-End Loss for Speaker Verification"
MIT License
349 stars 104 forks source link

noise dataset #3

Closed HaiFengZeng closed 5 years ago

HaiFengZeng commented 6 years ago

good job! Why should we use the noise dataset?

Janghyun1230 commented 5 years ago

Sorry for late reply. For text dependent verification we need several audio samples per person with same contents (words). However, it was hard to find suited public data so I generated them by adding different noises to certain word segment.