Noncompositional

Seeing Batman Begins yesterday, and in particular hearing Bale’s Batman voice, got me thinking about the ability to identify people based on various (combinations of) sensory stimuli. Take the situation where one person attempts to initiate a dialogue with another by using some typical verbal means, like calling out their name, or saying something like “hey,” or “uhh,” or what have you. Even aside from sensory input, we have several ways of determining or guessing who the caller might be, like knowledge of our surroundings and people in those surroundings, and who might be likely to call at a particular time. For instance, if it’s almost time for the 2:00 Thursday afternoon staff meeting, and you know that the guy in the next room is also going to the meeting, and it’s 2:05 and someone hucks a paper into your office from just outside reading, “Hey, better hurry to the meeting,” then you can make a pretty good guess as to who sent that message. And if instead of hucking a memo she tells you that you’re late (like a normal person would), then you’ve got even more clues. And of course if you can see the speaker, well there’s not much more your speaker identification system can ask for.

Okay, so how much would your coworker have to change her voice in order to baffle you as to who she is? What features would cause someone to have trouble with speaker identification? I suppose if I read some of the books on this list, or if I worked for one of these groups, I might have a better idea. But from a naive person’s point of view, I’d say that pitch, (local) accent, and spectral quality are some of the most important features. Some non-intuitive ones, nevertheless used by speaker recognition systems in the past, might be pause length, word length (speed), and venturing away from phonetic clues, word choice and grammatical constructions.

But more to the point, how much would someone gallivanting about in a shiny, squeaky bat-suit with a half-mask have to change their voice so that others wouldn’t be able to recognize them? Apparently, someone would have to speak in a raspy, low-pitched voice, “animalistic,” “savage,” voice (to quote some internet message board users) that is reminiscent of laryngitis, to repeat the comments of some friends who saw Returns with me. I was doubtful during the movie, but later reminisced on incidents where I was unable to identify who I was speaking to. In one particularly embarrassing incident, I called my office, where only I, a male coworker, and two female coworkers regularly work. A guy answered the phone, but it was slightly distorted, and I was completely unable to figure out who it was, despite the world knowledge that told me it nearly had to be my coworker. The situation quickly degraded as I misheard his “this is [name of coworker]” as “is this [name of coworker]?” and replied, “no, I want speak with [name].” I’m sure he was quite confused. Needless to say, I was quite confused, and I wasn’t even in danger of being thrown off the roof of a building, or facing a caped crusader.