I don't know if a reasonable analysis could be done without recording of Martin screaming assuming there are even recordings of him speaking.
I shudder to think it but if this case is tried it's possible they will have Zimmerman reenact screaming, record it from the same distance of the 911 call it was in the distant background on, and compare.
Well, if the question "is / is not" can be approached as one of elimination, one could use, oh, let's say ten test "no match" examples by reproducing the recording ten times with ten different subjects. Running each one through the analysis software with the same parameters used in the actual Zimmerman analysis should reveal a recognizable pattern.
Ten bucks says you will have one or possibly two tests that exceed one standard deviation of a >10% (guesstimate) match value. As I recall you would throw those out as that amount of "outliers" are to be expected.
The mean of the remaining tests will be your "elimination percentage." Compare that to the 48% value.
There is your real result.
< /quick-dirty >