公开(公告)号：US20240079019A1

公开(公告)日：2024-03-07

申请号：US18507824

申请日：2023-11-13

Applicant: Dolby Laboratories Licensing Corporation

Inventor： Roy M. FEJGIN ,  Grant A. DAVIDSON ,  Chih-Wei WU ,  Vivek KUMAR

IPC: G10L19/022 ,  G06F3/16 ,  G06N3/048 ,  G06N3/084

CPC classification number: G10L19/022 ,  G06F3/16 ,  G06N3/048 ,  G06N3/084

Abstract: Computer-implemented methods for training a neural network, as well as for implementing audio encoders and decoders via trained neural networks, are provided. The neural network may receive an input audio signal, generate an encoded audio signal and decode the encoded audio signal. A loss function generating module may receive the decoded audio signal and a ground truth audio signal, and may generate a loss function value corresponding to the decoded audio signal. Generating the loss function value may involve applying a psychoacoustic model. The neural network may be trained based on the loss function value. The training may involve updating at least one weight of the neural network.

公开(公告)号：US20180040336A1

公开(公告)日：2018-02-08

申请号：US15667359

申请日：2017-08-02

Applicant: Dolby Laboratories Licensing Corporation

Inventor： Chih-Wei WU ,  Mark S. VINTON

IPC: G10L21/0388 ,  G10L19/26

CPC classification number: G10L21/0388 ,  G10L19/26

Abstract: A system and method of blind bandwidth extension. The system selects a prediction model from a number of stored prediction models that were generated using an unsupervised clustering method (e.g., a k-means method) and a supervised regression process (e.g., a support vector machine), and extends the bandwidth of an input musical audio signal.

公开(公告)号：US20210082444A1

公开(公告)日：2021-03-18

申请号：US17046284

申请日：2019-04-10

Applicant: Dolby Laboratories Licensing Corporation

Inventor： Roy M. FEJGIN ,  Grant A. DAVIDSON ,  Chih-Wei WU ,  Vivek KUMAR

IPC: G10L19/022 ,  G06F3/16 ,  G06N3/08 ,  G06N3/04

Abstract: Computer-implemented methods for training a neural network, as well as for implementing audio encoders and decoders via trained neural networks, are provided. The neural network may receive an input audio signal, generate an encoded audio signal and decode the encoded audio signal. A loss function generating module may receive the decoded audio signal and a ground truth audio signal, and may generate a loss function value corresponding to the decoded audio signal. Generating the loss function value may involve applying a psychoacoustic model. The neural network may be trained based on the loss function value. The training may involve updating at least one weight of the neural network.