Speaker identification
Determining who is speaking by searching many enrolled voices (one-to-many).
Speaker identification determines who is speaking by comparing a voice against many enrolled voiceprints, a one-to-many search. Rather than checking a single claimed identity, it ranks the closest matches, which is how a system recognizes a returning caller without being told who they are.
Identification powers experiences like surfacing a customer’s history the moment they speak, even from a new phone number. It is more demanding than verification because the search space is larger, so calibrated scoring matters.
