Simultaneous dereverberation and denoising via low latency deep learning
US11404055B2 · kind B2 · utility
Assignee
Inventors
Key dates
| Filing date | Oct 16, 2020 |
| Grant date | Aug 2, 2022 |
| Priority date | — |
| Expiry date | Feb 3, 2041 |
Classification
- Technology area (CPC G)Physics
- CPC primaryG10L2021/02082
- WIPO fieldComputer technology
- WIPO sectorElectrical engineering
Abstract
A method includes extracting, from a frame of the audio sample, spectral features indicative of cochlear impulse response of an auditory system; obtaining an estimate of a speech signal in the frame from a neural network that is trained, during a training phase, to accept the spectral features as input and output the estimate of the speech signal, where the estimate of the speech signal includes reverberation and excludes noise present in the frame; mapping the estimate of the speech signal to a frequency domain of the frame using mapping parameters obtained during the training phase to obtain an estimate of a reverberant speech spectrum in the frame; and obtaining, from a time-distributed neural network, a dereverberant frame of the frame, where the estimate of the reverberant speech spectrum in the frame is used as an input to the time-distributed neural network.
Source: USPTO / EPO open patent data. Objective bibliographic and citation counts.