HPE IDOL Speech Server offers two sets of speaker identification tasks, based on different technologies. This section uses examples of the iVector-based tasks, but the original set of tasks are still available. The two sets of tasks have identical parameters and options unless otherwise specified. The previous tasks are maintained mainly for backwards compatibility, but in some cases might work better than new iVector tasks. The original tasks have same name as the iVector-based tasks, but without the Iv
prefix.
The speaker identification operation identifies sections of audio that contain speech from known speakers. HPE IDOL Speech Server lets you build a set of speaker templates to be identified based on audio samples that represent each speaker. This speaker set can be considered as either an open-set, or a closed-set:
|