BACK
X-Vectors
Compact numerical 'fingerprints' of a speaker's voice, used to tell speakers apart.
X-vectors are a type of speaker embedding: a compact numerical representation of the characteristics of a person's voice, produced by a neural network. They let systems compare and distinguish voices, powering speaker recognition, verification, and diarization. They are a widely used technique in speaker-aware speech processing.
Where this comes up at AudioShake