Intercepted calls are sent to S.P.I.D., which processes and scores them based on the probability they contain the target's voice.

Enrollment

Voiceprint enrollment requires a minimum of 30 seconds of net audio (target voice), preferably collected from two distinct sessions. When using stereo files, the system can automatically extract the specified side containing a target speaker. Mono files can be manually edited to retrieve the target voice using advanced Web-based tools. When enough audio is accumulated, the voiceprint is created and stored in a secure directory.

Verification

Verification and identification are performed by comparing audio files to a single voiceprint (verification) or a list of voiceprints (identification), detecting the participating speaker(s).