sounds like Bob Smith on line 42

This is speaker recognition, not voice recognition. It turns out that the problem is solved in a very different way for each; voice recognition, in order to be speaker independent, must throw out the information that makes it possible to do speaker recognition (and vice versa: the latter does not need some of the information that the former needs). This is used in "roving wiretaps" that apparently are used infrequently as they scan whole exchanges or number sets (e.g., all payphones in some city). The idea is to capture phone calls from the suspects that are savvy enough to know that they are being tapped. At least one Mob boss was caught this way (in Los Angeles, I think, about 4-5 years ago). Paul E. Baclace peb@procase.com