Signal Decoding with Sub-Band Noise Filling at Low Bit Rates

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing frequency domain codec algorithms face challenges in decoding signals due to insufficient bit allocation, leading to poor decoding quality, especially when bit rates are low, resulting in 'empty' or 'water-like' audio outputs as less important spectral coefficients are not encoded, severely affecting auditory quality.

Innovation Solution

A method and device for decoding signals that classify sub-bands into saturated and unsaturated bit allocation categories, with noise filling based on harmonic parameters to reconstruct spectral coefficients not obtained during decoding, improving signal quality by differentially performing noise filling.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If bits are allocated only to relatively important spectral coefficients during encoding, then the encoding efficiency is improved, but the decoding quality deteriorates because less important spectral coefficients are not encoded and cannot be reconstructed

Engineering Contradiction:
Improveencoding efficiencyVSAvoiddecoding quality
Core Design Contradiction:
ProductivityVSManufacturing precision

Solution Approach 1:

The patent applies copying by replicating decoded spectral coefficients to reconstruct undecoded spectral coefficients. Specifically, spectral coefficients that were not decoded due to bit allocation constraints are reconstructed by copying values from corresponding positions in previously decoded frames or adjacent frequency positions, thereby restoring the spectral information without requiring additional bits during encoding

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The patent changes the parameter of spectral coefficient reconstruction from simple zero-filling to intelligent replication based on temporal and spectral correlation. By analyzing the characteristics of decoded spectral coefficients and applying appropriate copying strategies (such as temporal replication from previous frames or spectral replication from adjacent bins), the system improves reconstruction quality while maintaining low bit rates

Inventive Principle:
Principle #35Parameter changes

2Manufacturing precision

If a noise filling method is used to reconstruct spectral coefficients by replicating decoded coefficients, then the decoding quality is improved compared to zero-filling, but the complexity of the decoding process increases

Engineering Contradiction:
Improvedecoding qualityVSAvoiddecoding process complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent segments the spectral band into different regions (decoded regions and undecoded regions) and applies different reconstruction strategies to each segment. By identifying which spectral coefficients were decoded and which were not, the system selectively applies copying operations only to the undecoded portions, thereby improving overall decoding quality while minimizing the additional processing complexity to only the necessary regions

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP3444817B1Method and device for decoding signal
Publication Date: 2021.07.21 HUAWEI TECH CO LTD
  • EP3444817B1 patent drawingFigure 1
  • EP3444817B1 patent drawingFigure 2~3
  • EP3444817B1 patent drawingFigure 4~5

AI summary

Embodiments of the present invention provide a method and device for decoding a signal. The method for decoding a signal includes: obtaining spectral coefficients of sub-bands from a received bitstream by means of decoding; classifying sub-bands in which the spectral coefficients are located into a sub-band with saturated bit allocation and a sub-band with unsaturated bit allocation; performing noise filling on a spectral coefficient that has not been obtained by means of decoding and is in the sub-band with unsaturated bit allocation, so as to reconstruct the spectral coefficient that has not been obtained by means of decoding; and obtaining a frequency domain signal according to the spectral coefficients obtained by means of decoding and the reconstructed spectral coefficient. In the foregoing embodiments of the present invention, a sub-band with unsaturated bit allocation in a frequency domain signal may be obtained by classification, and a spectral coefficient that has not been obtained by means of decoding and is in the sub-band with unsaturated bit allocation may be reconstructed, thereby improving signal decoding quality.