This does make use of multiple microphones and timing between sound waves arriving. To my understanding, the machine learning part allows it to work even as the

This does make use of multiple microphones and timing between sound waves arriving. To my understanding, the machine learning part allows it to work even as the speaker/wearer move relative to each other, such as if you turn your head.