System and method for stereophonic acoustic echo cancellation

摘要

Disclosed herein are systems, methods, and non-transitory computer-readable storage media for stereophonic acoustic echo cancellation. The method includes collecting, at a same time, a first audio sample of an audio source from a first omnidirectional microphone and a second audio sample of the audio source from a second omnidirectional microphone. The method includes delaying the second audio sample by a first amount of time to yield a delayed second audio sample and combining the delayed second audio sample with the first audio sample to produce a first channel, then delaying the first audio sample by a second amount of time to yield a delayed first audio sample and combining the delayed first audio sample with the second audio sample to produce a second channel. Then the method includes outputting the first channel and the second channel as a stereo audio signal of the audio source.

主权项

1. A method comprising:
collecting, at a same time, a first audio sample of an audio source from a first omnidirectional microphone and a second audio sample of the audio source from a second omnidirectional microphone; delaying the second audio sample by a first amount of time, to yield a delayed second audio sample; combining the delayed second audio sample with the first audio sample to produce a left channel {tilde over (X)}1(ω), wherein {tilde over (X)}1(ω)=S(ω)[eidk/2 cos φ1−e−i(φT1+kd/2 cos φ1], wherein d is a distance between the first omnidirectional microphone and the second omnidirectional microphone, wherein a first delay T1=d/c·cos φ1 is chosen to produce a back-to-back cardioid pattern that provides correlated signals for a direction-of-arrival from the audio source to a mid-point between the first omnidirectional microphone and the second omnidirectional microphone, and wherein c is a speed of sound; delaying the first audio sample by a second amount of time, to yield a delayed first audio sample, wherein the second amount of time is a distinct duration from the first amount of time; combining the delayed first audio sample with the second audio sample to produce a right channel {tilde over (X)}2 (ω), wherein {tilde over (X)}2(ω)=S(ω)[e−idk/2 cos φ2 −e−i(ωT2−kd/2 cos φ2], wherein a second delay T2=d/c·cos φ2 is chosen to produce the back-to-back cardioid pattern; and outputting the left channel and the right channel as a stereo audio signal of the audio source.