Patent #9,565,390 B1 was granted February 7, 2017, and it's the best thing to happen to mass surveillance since the people who've done nothing wrong filter.
From the summary: (full text PDF, SYSTEMS AND METHODS FOR ENHANCING RECORDED OR INTERCEPTED CALLS USING INFORMATION FROM A FACIAL RECOGNITION ENGINE)
A video stream is received. The video stream can be analyzed in real-time as it is being received or can be recorded and stored for later analysis. Information within the video streams can be extracted and processed by a facial and video content recognition engine and the information derived there from can be stored as metadata. The metadata can be queried for statistical data and/or for business or security analysis.
The metadata can be used to enrich the call content of a recorded or intercepted call. The information derived from the video streams can be used to determine whether or not the call should be recorded.
The inventor is Ofer Shochet of Tel Aviv and the patent is assigned to Verint Americas Inc.
Verint is an industry leader in call center analytics.
Obviously this kind of system has all sorts of uses for keeping America safe and stopping radical Islamic jihad terrors, and its potential is greater than the brief summary really indicates.
Drawing your attention one more time to the second paragraph from the patent:
"The metadata can be used to enrich the call content of a recorded or intercepted call. The information derived from the video streams can be used to determine whether or not the call should be recorded."
It can be assumed that Verint developed the technology for compatibility with more than one facial recognition engine (probably the three main algorithms, NEC, Morpho, and Cognitec). Even if only the major engines are integrated, all federal and most law enforcement facial recognition databases would be potential sources for Verint's system.