Resilient Vector Quantization for Frame-Erasure Audio Coding

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing audio codecs experience annoying artefacts due to prediction errors during frame erasure events, particularly after long runs of prediction in vector quantization, which affect the quality of decoded speech or audio signals.

Innovation Solution

A method and apparatus for quantizing line spectral frequency coefficients that switch between predictive and non-predictive modes based on a distortion measure comparison with a threshold, using a frame error concealment process to determine the optimal mode of operation for subsequent audio frames, thereby minimizing artefacts.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If predictive mode is used for quantizing LSF coefficients, then coding efficiency is improved, but artefacts occur during frame erasure events

Engineering Contradiction:
Improvecoding efficiencyVSAvoidartefacts in decoded signal
Core Design Contradiction:
ProductivityVSObject-affected harmful factors

Solution Approach 1:

The vector quantizer dynamically switches between predictive and non-predictive modes based on frame erasure detection. When frame erasure is detected, the system transitions from predictive mode to non-predictive mode to avoid propagating prediction errors that cause artefacts, while maintaining the ability to return to predictive mode when normal operation resumes.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The invention changes the operational parameter (prediction mode status) of the vector quantizer in response to frame erasure events. By modifying the mode parameter from predictive to non-predictive during erasure conditions, the system prevents the generation of harmful artefacts while preserving coding efficiency during normal operation.

Inventive Principle:
Principle #35Parameter changes

2Object-affected harmful factors

If non-predictive mode is used to avoid artefacts, then signal quality is improved, but coding efficiency decreases

Engineering Contradiction:
Improvesignal qualityVSAvoidcoding efficiency
Core Design Contradiction:
Object-affected harmful factorsVSProductivity

Solution Approach 1:

The system dynamically adjusts the quantizer mode based on operational conditions, using non-predictive mode only when necessary (during frame erasure) to maintain signal quality, and switching to predictive mode during normal operation to maximize coding efficiency. This dynamic adaptation resolves the trade-off by applying each mode selectively.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The operational parameter (mode selection) is changed conditionally based on frame erasure status. The system maintains high coding efficiency by default in predictive mode and only transitions to non-predictive mode when frame erasure is detected, thus preserving signal quality only when necessary without permanently sacrificing coding efficiency.

Inventive Principle:
Principle #35Parameter changes

3Quantity of substance

If predictive quantization is used continuously, then bit rate is reduced, but error propagation increases during frame erasure

Engineering Contradiction:
Improvebit rateVSAvoiderror propagation
Core Design Contradiction:
Quantity of substanceVSReliability

Solution Approach 1:

The system prepares for potential frame erasure by having a non-predictive mode ready to switch to. This pre-configured alternative prevents error propagation during frame erasure events without affecting normal predictive coding operation, thus cushioning against reliability issues while maintaining bit rate efficiency.

Inventive Principle:
Principle #11Beforehand cushioning (Prior cushioning)

Solution Approach 2:

The system changes the prediction parameter status in response to frame erasure detection. By transitioning from predictive to non-predictive mode during erasure events, the system prevents error propagation while maintaining low bit rate during normal operation, thus resolving the reliability-bit rate trade-off.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP2915166B1A method and apparatus for resilient vector quantization
Publication Date: 2018.10.17 NOKIA TECHNOLOGIES OY
  • EP2915166B1 patent drawingFigure 1~2
  • EP2915166B1 patent drawingFigure 3
  • EP2915166B1 patent drawingFigure 4

AI summary

It is inter alia disclosed to quantise a vector of a plurality of coefficients using a predictive mode of operation of a vector quantiser, wherein the vector quantiser can operate in either a predictive mode of operation or a non-predictive mode of operation, determine a vector of a plurality of recovered coefficients; compare the vector of the plurality of coefficients to the vector of the plurality of recovered coefficients; and determine the mode of operation of the vector quantiser for a vector of a plurality of coefficients associated with a subsequent frame of audio samples, wherein the mode of operation is dependent on the comparison.