US 12,456,482 B2
Neural temporal beamformer for noise reduction in single-channel audio signals
Saeed Mosayyebpour Kaskari, Irvine, CA (US)
Assigned to Synaptics Incorporated, San Jose, CA (US)
Filed by Synaptics Incorporated, San Jose, CA (US)
Filed on Jan. 26, 2023, as Appl. No. 18/160,278.
Prior Publication US 2024/0257827 A1, Aug. 1, 2024
Int. Cl. G10L 25/78 (2013.01); G10L 21/0208 (2013.01); G10L 25/93 (2013.01)
CPC G10L 25/78 (2013.01) [G10L 21/0208 (2013.01); G10L 25/93 (2013.01); G10L 2025/932 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A method of speech enhancement, comprising:
receiving a series of frames of a single-channel audio signal;
inferring a probability of speech (pDNN) in a first frame of the series of frames based on a neural network;
determining a voice activity detection (VAD) parameter based at least in part on the probability of speech pDNN, the VAD parameter indicating whether speech is present or absent in the first frame;
determining an interframe correlation (IFC) vector associated with a speech component of the audio signal based at least in part on the probability of speech pDNN and the VAD parameter, the IFC vector indicating an interframe correlation of the speech component between consecutive frames in the series of frames; and
filtering a noise component of the audio signal from the first frame based at least in part on the IFC vector and the series of frames.