Skip to content

Latest commit

 

History

History
5 lines (3 loc) · 764 Bytes

File metadata and controls

5 lines (3 loc) · 764 Bytes

Speaker Recognition

Speaker recognition models, also known as speaker verification or speaker identification models, are designed to identify and distinguish between different individuals based on their unique vocal characteristics. Here, they can be used both as a feature extraction component and for evaluation purposes (i.e. ASV evaluation).

Currently, we have implemented wrappers for the speaker recognition models of SpeechBrain. Configurations for all these models can be found under config/components/spkid. SpeechBrain has trained an x-vector model and an ECAPA-TDNN model on VoxCeleb 1&2.