Lattice Vector Quantizer Audio Encoder Complexity

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current transform coding approaches for audio signals improve quantization efficiency but do so at the cost of significant codec complexity.

Innovation Solution

The proposed solution involves evaluating quantization distortion in a transposed vector space to reduce encoding complexity, using techniques such as lattice vector quantization and sorting the input vector by absolute values.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If transform coding is used to improve quantization efficiency, then manufacturing precision is improved, but device complexity increases

Engineering Contradiction:
Improvequantization efficiencyVSAvoidcodec complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The audio signal is divided into multiple sub-bands using a filter bank, where each sub-band is processed independently with its own quantization parameters. This segmentation allows efficient quantization of different frequency regions while reducing the overall computational complexity by breaking down the complex full-band processing into simpler parallel sub-band operations.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Different quantization strategies and parameters are applied to different sub-bands based on their specific characteristics. Each sub-band uses locally optimized quantization parameters that match its spectral properties, improving overall quantization efficiency while keeping each local processing task simple and computationally manageable.

Inventive Principle:
Principle #3Local quality

2Manufacturing precision

If lattice vector quantization is used to improve quantization performance, then manufacturing precision is improved, but device complexity increases

Engineering Contradiction:
Improvequantization performanceVSAvoidencoding complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The vector quantization process is segmented into multiple stages: first dividing the audio signal into sub-bands, then extracting parameters from each sub-band, and finally applying lattice vector quantization to the parameter set. This multi-stage segmentation reduces the complexity of any single quantization operation while maintaining overall quantization performance.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The quantization approach transitions from direct time-domain or frequency-domain signal quantization to parameter-domain quantization. By extracting meaningful parameters from the audio signal and quantizing these parameters instead of the raw signal values, the system achieves efficient quantization with reduced computational complexity.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

Data Source

PatentEP3084761B9Audio signal encoder
Publication Date: 2025.06.11 NOKIA TECHNOLOGIES OY
  • EP3084761B9 patent drawingFigure 1
  • EP3084761B9 patent drawingFigure 2~3
  • EP3084761B9 patent drawingFigure 4

AI summary

An apparatus comprising: a vector generator configured to generate at least one vector of parameters defining at least one audio signal; a lattice vector quantizer configured to sort the at least one vector of parameters according to an ordering of at least one vector absolute tuples to generate an associated at least one ordered vector of parameters; the lattice vector quantizer configured to select from a list of leader classes at least one potential code vector; the lattice vector quantizer configured to determine a distance between the at least one potential code vector and the at least one ordered vector of parameters; the lattice vector quantizer configured to determine at least one leader class associated with a potential code vector which generates the smallest associated distance; the lattice vector quantizer configured to transpose the at least one leader class to generate an output lattice quantized codevector.