REAL-TIME SPEAKER IDENTIFICATION WITH DIARIZATION

摘要

The invention refers to a system for real-time identification of conference speaker identity, the system being configured to receive at least one input audio data stream representing an interaction between a plurality of individual speakers; wherein the system comprises a diarization unit for segmenting said one or more input audio data streams into speaker-homogenous audio segments; a speaker profile unit comprising a plurality of individual speaker profiles; and an identification unit for assigning said speaker-homogenous audio segments to an individual speaker; wherein the diarization unit is further configured to derive for an audio segment values for a second set of parameters; and wherein said system further comprises a controller which is configured for monitoring diarization, for controlling speaker identification of said segments, and for sharing conference speaker identity. The invention further refers to a method for adding and identifying an unidentified speaker to said system.