Multichannel Audio Noise Filling from Previous-Frame Spectral Lines

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing multichannel audio coding systems, such as Opus/Celt, MPEG-4 AAC, and MPEG-D xHE-AAC, face limitations in coding quality at very low bitrates for tonal or transient stereophonic inputs due to the need to explicitly transmit many spectral coefficients, especially when using noise filling and spectral band replication.

Innovation Solution

Implement inter-channel noise filling using spectrally co-located spectral lines from a previous frame, enhancing the noise filling process by utilizing the downmix of a previous frame as a source, and integrating this method into existing codecs like xHE-AAC in a semi-backward-compatible manner.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If noise filling and spectral band replication are used for low-bitrate coding, then coding efficiency is improved, but coding quality deteriorates for very tonal or transient stereophonic input

Engineering Contradiction:
Improvecoding efficiencyVSAvoidcoding quality
Core Design Contradiction:
ProductivityVSManufacturing precision

Solution Approach 1:

The patent copies spectral lines from a previous frame to fill zero-quantized scale factor bands in the current frame. This copying approach preserves tonal and transient characteristics better than artificial noise generation, while still achieving coding efficiency by avoiding explicit transmission of these coefficients.

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The patent performs preliminary action by storing and utilizing spectral lines from a previous frame before they are needed for noise filling. This allows the decoder to reconstruct current frame spectral lines using previously decoded information, improving quality without requiring additional transmission data.

Inventive Principle:
Principle #10Preliminary action

2Manufacturing precision

If many spectral coefficients are transmitted explicitly for tonal or transient input, then coding quality is improved, but bitrate consumption increases

Engineering Contradiction:
Improvecoding qualityVSAvoidbitrate consumption
Core Design Contradiction:
Manufacturing precisionVSQuantity of substance

Solution Approach 1:

Instead of transmitting spectral coefficients explicitly, the patent uses copying by deriving current frame spectral lines from previous frame spectral lines through simple operations (addition, subtraction, or copying), achieving high reconstruction quality at very low bitrates.

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The patent enables self-service by allowing the decoder to reconstruct spectral lines using only previously decoded information and simple arithmetic operations, without requiring additional transmitted data. The system uses its own previously decoded output to generate current frame spectral lines.

Inventive Principle:
Principle #25Self-service

3Quantity of substance

If zero-quantization is applied to scale factor bands, then bitrate is reduced, but reconstruction quality deteriorates

Engineering Contradiction:
ImprovebitrateVSAvoidreconstruction quality
Core Design Contradiction:
Quantity of substanceVSManufacturing precision

Solution Approach 1:

The patent applies copying by reconstructing zero-quantized spectral lines from previous frame spectral lines. This allows the decoder to fill in missing spectral information without transmitting additional data, maintaining reconstruction quality while achieving low bitrate through zero-quantization.

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The patent changes parameters by selectively applying zero-quantization to certain scale factor bands while using spectral line copying for others. This parameter-based approach allows flexible trade-off between bitrate and quality, applying aggressive compression only where spectral line copying can maintain quality.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP4369335B1Noise filling in multichannel audio coding
Publication Date: 2025.10.29 FRAUNHOFER GESELLSCHAFT ZUR FORDERUNG DER ANGEWANDTEN FORSCHUNG EV
  • EP4369335B1 patent drawingFigure 1
  • EP4369335B1 patent drawingFigure 2
  • EP4369335B1 patent drawingFigure 3

AI summary

In multichannel audio coding, an improved coding efficiency is achieved by the following measure: the noise filling of zero-quantized scale factor bands is performed using noise filling sources other than artificially generated noise or spectral replica. In particular, the coding efficiency in multichannel audio coding may be rendered more efficient by performing the noise filling based on noise generated using spectral lines from a previous frame of, or a different channel of the current frame of, the multichannel audio signal.