Noise Feedback Quantization for Audio Encoding

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Temporal Noise Shaping (TNS) in MPEG-2/4 AAC audio coding systems is limited by increased bit rate consumption, lack of backward compatibility, and increased decoder complexity due to the need for transmitting filter coefficients and applying inverse filters.

Innovation Solution

Noise Feedback Quantization (NFQ) is employed in digital audio encoding, where the quantization error is filtered and fed back as a signal to shape the noise in the frequency domain, allowing for dynamic control of noise shaping without requiring filter coefficient transmission to the decoder.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If Temporal Noise Shaping (TNS) is used to control quantization noise temporal envelope, then noise shaping performance is improved, but bit rate consumption increases due to filter coefficient transmission

Engineering Contradiction:
Improvenoise shaping performanceVSAvoidbit rate consumption
Core Design Contradiction:
Manufacturing precisionVSLoss of information

Solution Approach 1:

The patent extracts the noise shaping function from the traditional TNS framework that requires filter coefficient transmission. By using a modified quantization process that inherently shapes noise without requiring separate filter parameters, the solution removes the problematic element (filter coefficient transmission) while preserving the beneficial effect (noise shaping performance).

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The quantization process itself is made to perform the noise shaping function without requiring additional components or parameters. The quantizer is designed to automatically shape the temporal envelope of quantization noise through its internal operation, making the system self-sufficient and eliminating the need for external filter coefficient management.

Inventive Principle:
Principle #25Self-service

2Manufacturing precision

If TNS filter coefficients are transmitted to the decoder, then noise shaping control is improved, but backward compatibility is lost

Engineering Contradiction:
Improvenoise shaping controlVSAvoidbackward compatibility
Core Design Contradiction:
Manufacturing precisionVSAdaptability or versatility

Solution Approach 1:

The patent removes the requirement for filter coefficient transmission that breaks backward compatibility. By redesigning the noise shaping mechanism to not depend on transmitted filter parameters, the solution maintains compatibility with legacy decoders that expect standard quantization formats while achieving improved noise shaping control through the modified quantization process.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The modified quantization process serves multiple functions simultaneously: it performs standard quantization for compatibility with existing decoders while also implementing noise shaping control. This multi-functionality allows the same encoding process to benefit both legacy and advanced systems without requiring separate processing paths.

Inventive Principle:
Principle #6Universality (Multi-functionality)

3Manufacturing precision

If inverse filter is applied in the decoder, then noise shaping performance is improved, but decoder complexity increases

Engineering Contradiction:
Improvenoise shaping performanceVSAvoiddecoder complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent extracts the noise shaping function from the decoder side operations. By moving the noise shaping control to the encoder's quantization process, the solution removes the need for complex inverse filtering operations in the decoder, thereby reducing decoder complexity while preserving noise shaping performance through the encoded signal characteristics.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The encoded signal itself carries the noise shaping information without requiring additional decoder processing. The quantization process embeds the temporal envelope control directly into the quantized coefficients, allowing the decoder to reconstruct the shaped noise characteristics through simple inverse transformation without requiring complex filtering operations.

Inventive Principle:
Principle #25Self-service

4Manufacturing precision

If filter parameters are transmitted as side information, then temporal envelope control is improved, but bit rate efficiency decreases

Engineering Contradiction:
Improvetemporal envelope controlVSAvoidbit rate efficiency
Core Design Contradiction:
Manufacturing precisionVSProductivity

Solution Approach 1:

The patent removes the side information transmission requirement for temporal envelope control. By integrating the noise shaping control directly into the quantization process parameters that are already transmitted for standard decoding, the solution achieves temporal envelope control without requiring additional bit rate allocation for filter parameters.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The existing quantization parameters serve dual purposes: they control both the standard quantization process and the temporal envelope of quantization noise. This multi-functionality allows the same parameter set to achieve both compression and noise shaping objectives, eliminating the need for separate side information channels and improving bit rate efficiency.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentUS8706507B2Arbitrary shaping of temporal noise envelope without side-information utilizing unchanged quantization
Publication Date: 2014.04.22 DOLBY LABORATORIES LICENSING CORP
  • US8706507B2 patent drawing
  • US8706507B2 patent drawing
  • US8706507B2 patent drawing

AI summary

In a first aspect, arbitrary shaping of the temporal envelope of noise is provided in spectral domain coding systems without the need of side-information. In the encoding, a filtered measure of quantization error is applied as a feedback signal to the frequency-domain representation of a discrete time-domain signal prior to quantization, so that the filtering parameters of said filtering affect the shaping of quantization noise in the time domain of the quantized frequency-domain representation with unchanged quantization of the discrete time-domain signal when it is inversely transformed from the frequency domain back to the time domain in decoding. This may be accomplished with respect to each of a plurality of frequency bins or groups of bins. In another aspect, frequency-domain noise-feedback quantizing in digital audio encoding is provided.