[mikewarot]

For actual speakers, you could use a second of the same type of speaker as a microphone, and assume the losses are similar, do a slow sine sweep through, and just divide all the dB values in half.

Obviously this will only work if there's a passive crossover network, and nothing active in the speakers.