I'm not certain as to why it would have to have a flat response curve. My impression is that it is only necessary to pick up voice. So if it is tuned (for lack of a better word) to human vocal frequencies, wouldn't that be good enough? My first impression is that it would be better that way so as to "cut out" all of the non-voice frequencies that may confuse the voice recognition system.