US 12,272,369 B1
Dereverberation and noise reduction
Amit Singh Chhetri, Sunnvale, CA (US); Mrudula V. Athi, Natick, MA (US); Pradeep Kumar Govindaraju, San Jose, CA (US); and Rong Hu, San Jose, CA (US)
Assigned to Amazon Technologies, Inc., Seattle, WA (US)
Filed by Amazon Technologies, Inc., Seattle, WA (US)
Filed on Jan. 19, 2022, as Appl. No. 17/578,737.
Int. Cl. G10L 21/0216 (2013.01); G10L 21/0208 (2013.01); G10L 21/10 (2013.01); G10L 25/30 (2013.01)
CPC G10L 21/0216 (2013.01) [G10L 21/10 (2013.01); G10L 25/30 (2013.01); G10L 2021/02082 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method, the method comprising:
receiving first audio data corresponding to first audio generated by a loudspeaker;
receiving second audio data corresponding to audio captured by at least one microphone, the second audio data representing the first audio and second audio, the second audio including speech;
performing, using the first audio data and the second audio data, echo cancellation to generate third audio data and echo estimate data;
generating, using a neural network and the third audio data, first data representing the second audio, wherein the neural network is configured to perform noise reduction and dereverberation processing configured to reduce a reverberation associated with the speech; and
generating, using the first data, fourth audio data representing the second audio.