Model-Based Subband Prediction in Critically Sampled Filterbanks

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing audio source coding systems face challenges in efficiently encoding subband predictors at low bit rates while minimizing noise shaping artifacts and aliasing artifacts, particularly in critically sampled filterbanks used for low bit rate audio coding.

Innovation Solution

Implementing a compact description of subband predictors based on signal models, using cross-subband predictor terms, and directly applying predictors in the subband domain to reduce computational complexity and aliasing artifacts.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of information

If backwards adaptive prediction scheme is used to avoid transmitting predictor description, then bit rate is reduced, but prediction quality degrades significantly at low bit rates

Engineering Contradiction:

Improvebit rateVSAvoidprediction quality

Core Design Contradiction:

Loss of informationVSReliability

Solution Approach 1:

The system uses previously decoded portions of the audio signal to estimate and determine the predictor parameters at both encoder and decoder without transmission. The decoder autonomously reconstructs the predictor by analyzing the decoded signal history, eliminating the need for explicit predictor description transmission while maintaining prediction quality through self-adaptive parameter estimation.

Inventive Principle:

Principle #25Self-service

2Manufacturing precision

If subband prediction is applied in critically sampled filterbanks, then noise shaping is improved, but aliasing artifacts appear in the output audio signal

Engineering Contradiction:

Improvenoise shaping controlVSAvoidaliasing artifacts

Core Design Contradiction:

Manufacturing precisionVSObject-generated harmful factors

Solution Approach 1:

The system extracts and removes aliasing artifacts from the prediction error signal by applying anti-aliasing filtering operations. The filterbank structure is modified to include specific filtering stages that eliminate the harmful aliasing components while preserving the beneficial noise shaping characteristics of subband prediction.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The system converts the aliasing artifacts generated by subband prediction into beneficial components by using them to inform the predictor design. The aliasing patterns are analyzed and used to adjust predictor parameters, transforming the harmful artifacts into useful information for improving prediction accuracy and noise shaping.

Inventive Principle:

Principle #22Blessing in disguise (Convert harm into benefit)

3Reliability

If long term predictor is implemented with large delay in time domain, then periodic signal prediction is improved, but computational complexity increases due to additional filterbank operations

Engineering Contradiction:

Improveperiodic signal predictionVSAvoidcomputational complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

Instead of implementing the long term predictor in the time domain with large delay and then converting to subband domain, the system inverts the approach by directly implementing the predictor in the subband domain. This reversal eliminates the need for additional filterbank conversion operations while maintaining the ability to predict periodic signals effectively through subband-specific predictor parameters.

Inventive Principle:

Principle #13The other way round (Inversion)

Solution Approach 2:

The system replaces the mechanical time-domain delay and filterbank conversion mechanism with a direct subband-domain prediction mechanism. By substituting the complex time-domain implementation with a frequency-domain approach, the system achieves equivalent periodic signal prediction without the computational overhead of additional filterbank operations.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS20250285629A1Model Based Prediction in a Critically Sampled Filterbank

Publication Date: 2025.09.11 DOLBY INTERNATIONAL AB

AI summary

The present document relates to audio source coding systems which make use of linear prediction in combination with a filterbank. A method for estimating a sample of a subband signal from two or more previous samples of the subband signal is described. The subband signal corresponds to a plurality of subbands, having an equal subband spacing, of a subband-domain representation of an audio signal. The method comprises determining signal model data using a model parameter; determining a first prediction coefficient in response to the model parameter using a first lookup table and/or a first analytical function; determining a second prediction coefficient in response to the model parameter using a second lookup table and/or a second analytical function; and determining the estimate of the sample by applying the first prediction coefficient to the first previous sample and applying the second prediction coefficient to the second previous sample.