A headset and a method for determining that a headset user is speaking, includes receiving a first signal from a first microphone and receiving a second signal from a second microphone. A principal signal is provided from a sum of the first and second signals. A reference signal is provided from a difference between the first and second signals. A rate of change of at least one of the principal signal or the reference signal is limited by a time constant. The principal signal and the reference signal are compared, and an indication that the user is speaking is selectively made based at least in part upon the comparison.