An apparatus for selectively extracting components of an input signal comprises a plurality of microphones for detecting an audio signal in a plurality of spaced locations. A plurality of signal samplers sample each spaced version of the audio signal. A microphone pattern processor derives a reference signal, a null signal and a core signal from the sampled signals. An adjustment frequency spectrum generator coupled to the null and reference signals generates an adjustment frequency spectrum signal that is applied to the core signal. An audio signal regenerator provides an output audio signal from the adjusted core signal. The apparatus can be used in a wireless device to pick up a speaker's voice in a noisy environment.