A facemask communication system includes a plurality of microphones within the inside compartment of the facemask for detecting sounds voiced by the wearer. The microphones have cardioid sound reception patterns even when the region at the rear of the microphones are blocked by channeling sound to the rear surfaces of the diaphragms through a sound guide. The signals from the microphones are wirelessly transmitted through the transparent window of the facemask using electromagnetic radiation. After transmission through the transparent window, signal processing is performed on the signals. A digital signal processor processes the audio signal generated by the microphones in the time and frequency domain. The signal processor can include at least one adaptive filter to enhance a wanted signal in the audio signal and at least one adaptive filter to reduce an unwanted signal in the audio signal. Breathing noise in the signals is reduced using a plurality of microphones to provide signals to the signal processor. The system provides voice recognition for automatically responding to commands.