Basic Idea

If we assume the talker is far away from the microphone array, the acoustic wave will be nearly planer. Notice that the wavefront propogates across every microphone. The simplest way to make use of this is to delay the signal received at each microphone and sum them together. This has the effect of increasing the Signal to Noise Ratio (SNR) and increasing the Signal to Interference plus Noise Ratio (SINR) if the interference is at a different angle than the source relative to the array.

In the example below, consider the large squiggle to be our source, the little squiggle to be interference, and the fuzziness to be noise. If S1 and S2 are delayed in time and added to S3, the result is the bottom signal. Notice the noise and interference seem smaller. (A real signal would not have the source and interference separated in time. And the first and last signal would be seperated by at most a few samples.)