In terms of speech recognition
two microphones
are better than one, and sound waves
reaching multiple
microphones
can exhibit varying degrees
of time delay, which can be used to enhance
the intensity of signals from a particular direction while reducing signals from other directions.
Researchers at the Amazon Alexa division have developed a novel acoustic modeling
framework that enhances performance through unified speech enhancement and speech
recognition. They used a dual-microphone system in their experiments, compared to systems
that used seven microphones. They said the new system reduced the error rate of speech
recognition by 9.5%.