Voice 'Fingerprints' Change Crime-Solving

I don't know what these fellows have been doing, evidently research is not their forte.

Both the earlier Echelon and the later Carnivore Promis systems have voice recognition, and have had voice recognition for years.

OK, so anywhere in the world you make a phone call, it it is automatically recorded, the computers spin, and your calling or called location is immediately noted, because your voice-print has been on-file for many years.

It's better that you know the facts, rather than believe some journalistic "flannel".

Kind Regards....

Sparkstation--

I don't think you have it quite right. Echelon, et al. (as I understand it) is set to process a voice stream in search of a "hit" word from a list of hit words. Depending on the number of hits in a stream or source, the recording is then referred to a live human being for further analysis. The big advance in these systems was to be able to detect a given word in spite of the speaker's gender, accent, rate of speech, vocal pitch and background noise. These programs are used as a fast filter for massive amounts of traffic. Only the final few products are analyzed for content or source by intel analysts. Even then there are problems finding enough humans to do the final analysis. Rumor has it that there was a large increase in chatter that was picked up by these programs just prior to 9/11, but it wasn't until after a week or so that the relatively few Arab speaking analysts caught up with the review backlog.

Voice fingerprints depend on averaging a spectral analysis and profile of a subject's enunciation of a given word. A lot depends on the quality of the source recording. The advance may be that there has been a breakthrough in being able to derive the fingerprint spectrum independent of word or sound quality.