Quantization Unit for Voice Codec SNR Adaptation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Transform-based audio codec systems, such as AAC, degrade faster in quality for voice signals compared to musical signals at lower bitrates, especially for dry speech signals, due to limitations in quantization schemes used in these systems.

Innovation Solution

A quantization unit is introduced that utilizes a set of quantizers with different signal-to-noise ratios (SNRs) to improve the quality of voice signals, including noise-filling, dithered, and un-dithered quantizers, which are ordered to provide incremental SNR improvements, allowing for efficient bit allocation and reduced audible artifacts.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of information

If transform-based audio codec systems (e.g., AAC) use conventional quantization schemes, then coding gain is achieved, but perceptual quality of voice signals degrades faster at lower bitrates compared to musical signals

Engineering Contradiction:
Improvecoding gainVSAvoidperceptual quality of voice signals
Core Design Contradiction:
Loss of informationVSManufacturing precision

Solution Approach 1:

The patent applies parameter changes by transitioning from conventional scalar quantization to vector quantization methods. Specifically, it uses Twin Vector Quantization (TVQ) and Multi-stage Vector Quantization (MSVQ) which change the quantization parameter structure from scalar to vector, enabling better preservation of voice signal quality at lower bitrates while maintaining coding efficiency

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent employs composite quantization schemes that combine multiple quantization stages and methods. The Twin Vector Quantization combines two separate vector quantizers, while Multi-stage Vector Quantization uses multiple sequential quantization stages, creating a composite approach that achieves both high coding gain and preserved perceptual quality

Inventive Principle:
Principle #40Composite materials

2Device complexity

If conventional scalar quantization is used in transform-based audio coding, then operational complexity is reduced, but spectral holes and audible artifacts increase in voice signals

Engineering Contradiction:
Improveoperational complexityVSAvoidspectral holes and audible artifacts
Core Design Contradiction:
Device complexityVSObject-generated harmful factors

Solution Approach 1:

The patent uses composite quantization structures combining multiple vector quantizers or multiple quantization stages. Twin Vector Quantization employs two parallel vector quantizers that work together to fill spectral holes, while Multi-stage Vector Quantization uses sequential stages that progressively refine the quantization, both reducing audible artifacts while managing complexity

Inventive Principle:
Principle #40Composite materials

Solution Approach 2:

The patent applies copying principles by using multiple quantizers (in TVQ) or multiple stages (in MSVQ) that create redundant representations of the signal. This redundancy allows the system to fill spectral holes and reduce artifacts by distributing the quantization burden across multiple copies or stages

Inventive Principle:
Principle #26Copying

Data Source

PatentEP3217398B1Advanced quantizer
Publication Date: 2019.08.14 DOLBY INTERNATIONAL AB
  • EP3217398B1 patent drawingFigure 1A
  • EP3217398B1 patent drawingFigure 1B
  • EP3217398B1 patent drawingFigure 2~3A

AI summary

The present document relates an audio encoding and decoding system (referred to as an audio codec system). In particular, the present document relates to a transform-based audio codec system which is particularly well suited for voice encoding/decoding. A quantization unit (112) configured to quantize a first coefficient of a block (141) of coefficients is described. The block (141) of coefficients comprises a plurality of coefficients for a plurality of corresponding frequency bins (301). The quantization unit (112) is configured to provide a set (326, 327) of quantizers. The set (326, 327) of quantizers comprises a plurality of different quantizers (321, 322, 323) associated with a plurality of different signal-to-noise ratios, referred to as SNR, respectively. The plurality of different quantizers (321, 322, 323) includes a noise-filling quantizer (321); one or more dithered quantizers (322); and one or more undithered quantizers (323). The quantization unit (112) is further configured to determine an SNR indication indicative of a SNR attributed to the first coefficient, and to select a first quantizer from the set (326, 327) of quantizers, based on the SNR indication. In addition, the quantization unit (112) is configured to quantize the first coefficient using the first quantizer.