Mere hours after Operation Iraqi Freedom commenced, a bespectacled “Saddam Hussein” appeared on Iraqi TV to deliver a pep talk. Some U.S. officials suspect that the speaker was not really Saddam, but rather one of the dictator’s many body doubles. How can the CIA figure out whether we’ve been duped?
They’ll start by analyzing the speaker’s voice, much in the same way the National Security Agency has authenticated audiocassettes purported to contain statements by Osama Bin Laden. The CIA will begin by building a digital model of Saddam’s voice, using archived snippets from past speeches. Samples from last night’s address will then be fed into an automated speaker recognition system, which will compare the speaker’s sonic characteristics with those of Saddam. If the amplitude and frequency of the two voices are similar enough, the software will conclude that the Iraqi president was, indeed, the man who urged the nation to pick up its swords.
Another technological option is facial-recognition software, which analyzes such physical traits as nose slope and brow height. Such a program could compare the speaker’s visage with verifiable images of Saddam and make a determination as to whether they’re one and the same. A too-broad mouth or saggy cheek would be a tip off that Saddam sent an imposter in his stead.
Voice analysis and facial recognition are far from foolproof, though, especially given the poor audio and visual quality of the last night’s broadcast. Automated speaker recognition systems are primarily designed to handle “clean” samples, rather than muddy signals harvested from over-the-air sources. As for facial recognition, its track record is decidedly dicey; a Federal Aviation Administration test last May revealed a 53 percent error rate in one system. Those clunky glasses donned by “Saddam” certainly won’t help matters, as they obfuscate much of the speaker’s mid-facial region.
In the end, the CIA may use technology in combination with old-fashioned observation. One option is to cleanse the tape of background noise using audio-editing software, and then to have Arabic-speaking intelligence experts listen carefully. (It’s been said that Saddam’s lisp is fairly hard to imitate correctly.) And another low-tech standby, freeze frame, can be used to analyze whether the speaker looks a little thin in the shoulders, or is sporting a mustache that’s a shade too gray. They’ll never be 100 percent sure one way or the other, but the CIA should have a rough guess prepared soon.
Explainer thanks Lawrence Hornak of West Virginia University and William Knowles of c4i.org.
