Machine learning-based key generation for key-guided audio signal transformation

Patent metadata

Field	Value
Publication number	US-12567426-B2
Application number	US-202318104047-A
Country	US
Kind code	B2
Filing date	Jan 31, 2023
Priority date	Jul 31, 2020
Publication date	Mar 3, 2026
Grant date	Mar 3, 2026

Abstract

A method comprise: receiving input audio and target audio having a target audio characteristic; using a first neural network, trained to generate key parameters that represent the target audio characteristic based on one or more of the target audio and the input audio, generating the key parameters; and configuring a second neural network, trained to be configured by the key parameters, with the key parameters to cause the second neural network to perform a signal transformation of the input audio, to produce output audio having an output audio characteristic corresponding to and that matches the target audio characteristic.

First claim

What is claimed is:

1 . A method comprising:
receiving input audio and independent target audio having a target audio characteristic; using a first neural network deployed at a radio transmitter, trained to generate key parameters that represent the target audio characteristic based on one or more of the target audio and the input audio, generating the key parameters based on one or more of the target audio and the input audio; encoding the input audio into encoded input audio at the radio transmitter; multiplexing the key parameters and the encoded input audio at the radio transmitter; transmitting the multiplexed key parameters and the multiplexed encoded input audio from the radio transmitter; receiving the multiplexed key parameters and the encoded input audio at a radio receiver; demultiplexing the multiplexed key parameters and the multiplexed encoded input audio at the radio receiver; decoding the demultiplexed encoded input audio to recover the input audio at the radio receiver; and configuring, at the radio receiver, a second neural network deployed at the radio receiver to perform a signal transformation of audio representative of the decoded input audio using the demultiplexed key parameters to produce output audio having an output audio characteristic that matches the target audio characteristic.

2 . The method of claim 1 , wherein:
the generating the key parameters with the first neural network includes generating the key parameters to represent the target audio characteristic of the target audio.

3 . The method of claim 1 , wherein:
the input audio and the target audio include respective sequences of audio frames; the generating the key parameters includes generating the key parameters on a frame-by-frame basis; and the configuring the second neural network includes configuring the second neural network with key parameters generated on a frame-by-frame basis to cause the second neural network to perform the signal transformation on the frame-by-frame basis, to produce the output audio as a sequence of audio frames.

4 . The method of claim 1 , wherein:
the first neural network and the second neural network were trained jointly to minimize a combined cost, including a first cost associated with the key parameters and a second cost associated with the signal transformation.

5 . The method of claim 4 , wherein the combined cost is configured to drive back propagation of cost gradients of the combined cost with respect to each of the first neural network and the second neural network.

6 . The method of claim 4 , wherein:
the first cost measures mutual orthogonality between vectors representative of the key parameters; and the second cost represents an error between training target audio and training output audio produced by the signal transformation.

7 . The method of claim 1 , wherein receiving the
input audio comprises: receiving a bit-stream including theencoded input audio; and
decoding the encoded input audio to recover the input audio.

8 . The method of claim 7 , wherein:
the first neural network is trained to generate the key parameters to approximate target key parameters derived algorithmically from the target audio, such that the key parameters minimize an error between the key parameters and the target key parameters.

9 . The method of claim 7 , wherein:
the generating the key parameters using the first neural network includes generating the key parameters as spectral envelope key parameters including linear prediction (LP) coefficients (LPCs) or line spectral frequencies (LSFs) that represent a target spectral envelope of the target audio; and the configuring includes configuring the second neural network with the spectral envelope key parameters to cause the second neural network to perform the signal transformation as a transformation of an input spectral envelope of the input audio to an output spectral envelope of the output audio that matches the target spectral envelope.

10 . The method of claim 7 , wherein:
the generating the key parameters includes generating the key parameters as harmonic key parameters that represent target harmonics present in the target audio; and the configuring includes configuring the second neural network with the harmonic key parameters to cause the second neural network to perform the signal transformation of the audio representative of the input audio, such that the output audio includes harmonics that match the target harmonics.

11 . The method of claim 7 , wherein:
the generating the key parameters includes generating the key parameters as temporal key parameters that represent a target temporal characteristic of the target audio; and the configuring includes configuring the second neural network with the temporal key parameters to cause the second neural network to perform the signal transformation as a transformation of an input temporal characteristic of the input audio to an output temporal characteristic of the output audio that matches the target temporal characteristic.

12 . A system comprising:
a transmitter including a radio coupled to a processor and configured to: receive input audio and independent target audio having a target audio characteristic; use a first neural network, trained to generate key parameters that represent the target audio characteristic based on one or more of the target audio and the input audio, to generate the key parameters based on one or more of the target audio and the input audio; encode the input audio into encoded input audio; and multiplex the key parameters and the encoded input audio at the radio transmitter; and transmit the multiplexed key parameters and the multiplexed encoded input audio; and a receiver including a radio coupled to a processor and configured to:
receive the multiplexed key parameters and the encoded input audio;
demultiplex the multiplexed key parameters and the multiplexed encoded input audio;
decode the demultiplexed encoded input audio to recover the input audio; and
configure a second neural network, to perform a signal transformation of audio representative of the decoded input audio using the demultiplexed key parameters to produce output audio having an output audio characteristic that matches the target audio characteristic.

13 . The system of claim 12 , wherein:
the first neural network is configured to generate the key parameters to represent a target characteristic of the target audio.

14 . The system of claim 12 , wherein:
the input audio and the target audio include respective sequences of audio frames;
the first neural network is configured to generate the key parameters by generating the key parameters on a frame-by-frame basis; and
the receiver is configured to configure the second neural network by configuring the second neural network with key parameters generated on the frame-by-frame basis to cause the second neural network to perform the signal transformation on a frame-by-frame basis, to produce the output audio as a sequence of audio frames.

15 . The system of claim 12 , wherein:
the first neural network and the second neural network were trained jointly to minimize a combined cost, including a first cost associated with the key parameters and a second cost associated with the signal transformation.

16 . The system of claim 15 , wherein the combined cost is configured to drive back propagation of cost gradients of the combined cost with respect to trainable model parameters of each of the first neural network and the second neural network.

Assignees

• Dts Inc

Inventors

• Fejzo Zoran
• Kalker Antonius
• Venkatraman Atti

Classifications

• G10L25/30 Primary

  using neural networks · CPC title

• G06N3/045

  Combinations of networks · CPC title

• G06N3/084 Primary

  Backpropagation, e.g. using gradient descent · CPC title

• G06N3/044

  Recurrent networks, e.g. Hopfield networks · CPC title

• G10L21/0316

  by changing the amplitude · CPC title