Recurrent-Layer Vocoder Reconstruction for Low-Bitrate Audio

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing vocoder techniques struggle to efficiently generate high-quality audio signals from bitstreams, particularly at low bit rates, due to limitations in processing and conditioning methods.

Innovation Solution

An audio generator system utilizing a combination of preconditioning and conditioning learnable layers, including recurrent layers, to process audio signals in a sequence of frames, enhancing the generation of audio signals by applying conditioning feature parameters.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If conventional vocoder techniques are used to generate audio signals from bitstreams, then the system complexity is reduced, but the audio quality deteriorates particularly at low bit rates

Engineering Contradiction:
Improveaudio qualityVSAvoidsystem complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The audio generator is divided into multiple specialized processing blocks: a first processing block that receives bitstream data, a second processing block that receives input signal data, and a styling element that combines them. This segmentation allows each block to be optimized for its specific function, improving overall audio quality while managing system complexity through modular design.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The styling element acts as an intermediary component that receives processed data from both the first processing block (bitstream) and the second processing block (input signal), and combines them to produce the final audio output. This intermediary structure enables sophisticated audio reconstruction at low bit rates by leveraging both compressed data and original signal characteristics.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Productivity

If recurrent learnable layers are added to the preconditioning layer, then the processing capability is improved, but the device complexity increases

Engineering Contradiction:
Improveprocessing capabilityVSAvoiddevice complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The recurrent learnable layers in the first processing block perform preliminary processing on the bitstream data before it reaches the styling element. By pre-processing the compressed audio data with sophisticated recurrent networks, the system prepares enhanced feature representations that improve final audio quality while allowing the rest of the system to remain relatively simple.

Inventive Principle:
Principle #10Preliminary action

3Loss of substance

If low bit rate encoding is used, then the data transmission efficiency is improved, but the audio quality deteriorates

Engineering Contradiction:
Improvedata transmission efficiencyVSAvoidaudio quality
Core Design Contradiction:
Loss of substanceVSManufacturing precision

Solution Approach 1:

The system changes the parameter of bit rate to low values to achieve efficient data transmission, while compensating for the quality loss through the use of recurrent learnable layers and the styling element that intelligently reconstructs audio signals by combining compressed data with input signal characteristics.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The audio output is formed as a composite by combining processed bitstream data from the first processing block with processed input signal data from the second processing block through the styling element. This composite approach allows the system to achieve high audio quality despite using low bit rate encoding by leveraging multiple data sources.

Inventive Principle:
Principle #40Composite materials

Data Source

PatentEP4494137B1Vocoder techniques
Publication Date: 2025.10.15 FRAUNHOFER GESELLSCHAFT ZUR FORDERUNG DER ANGEWANDTEN FORSCHUNG EV
  • EP4494137B1 patent drawingFigure 1a
  • EP4494137B1 patent drawingFigure 1b
  • EP4494137B1 patent drawingFigure 1c

AI summary

There is disclosed and audio generator (10), configured to generate an audio signal (16) from a bitstream (3), the bitstream (3) representing the audio signal (16), the audio signal being subdivided in a sequence of frames, the audio generator (10) comprising: a first data provisioner (702) configured to provide, for a given frame, first data (15) derived from an input signal (14); a first processing block (40, 50, 50a-50h), configured, for the given frame, to receive the first data (15) and to output first output data (69) in the given frame, wherein the first processing block (50) comprises: at least one preconditioning learnable layer (710) configured to receive the bitstream (3), or a processed version (112) thereof, and, for the given frame, output target data (12) representing the audio signal (16) in the given frame; at least one conditioning learnable layer (71, 72, 73) configured, for the given frame, to process the target data (12) to obtain conditioning feature parameters (74, 75) for the given frame; and a styling element (77), configured to apply the conditioning feature parameters (74, 75) to the first data (15, 59a) or normalized first data (59, 76'); wherein the at least one preconditioning learnable layer (710) includes at least one recurrent learnable layer.