Audio Codec Post-Filter for Noise Reduction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional speech codecs face challenges in maintaining high-quality decoded speech due to noise and artifacts introduced by lossy compression, and existing post-processing techniques do not adequately address these issues.

Innovation Solution

The implementation of advanced filtering techniques, including the calculation and application of specific filter coefficients in the frequency domain to enhance the quality of reconstructed audio signals, particularly by attenuating noise in spectral valleys and enhancing energy in frequency regions near band intersections.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of energy

If lossy compression is used to reduce bit rate, then storage and transmission cost decreases, but noise and artifacts are introduced in decoded speech

Engineering Contradiction:
Improvebit rateVSAvoidnoise and artifacts
Core Design Contradiction:
Loss of energyVSObject-affected harmful factors

Solution Approach 1:

The patent applies post-filtering to convert the harmful noise and artifacts introduced by lossy compression into beneficial effects. By analyzing the spectral characteristics of the decoded speech and applying frequency-domain filtering, the system attenuates noise in spectral valleys while preserving useful signal components, thereby improving perceived quality despite the underlying compression losses

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

Solution Approach 2:

The system dynamically adjusts filtering parameters based on the spectral characteristics of the decoded speech. By computing filter coefficients that adapt to the specific frequency content and noise distribution in each frame, the post-filter optimizes the balance between noise reduction and signal preservation, transforming the fixed compression parameters into adaptive quality enhancement

Inventive Principle:
Principle #35Parameter changes

2Object-affected harmful factors

If conventional post-filters are applied to reduce noise, then some quality improvement is achieved, but existing techniques do not adequately address noise in spectral valleys and energy loss at band intersections

Engineering Contradiction:
ImprovenoiseVSAvoidquality of decoded speech
Core Design Contradiction:
Object-affected harmful factorsVSManufacturing precision

Solution Approach 1:

The patent implements frequency-selective filtering that applies different processing characteristics to different frequency regions. Specifically, the filter attenuates noise in spectral valleys (low energy regions) while preserving energy in formant regions, and applies special handling to frequency regions near band intersections where energy is typically lost during compression. This localized quality enhancement addresses the specific weaknesses of conventional uniform post-filters

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The frequency spectrum is segmented into distinct regions (spectral valleys, formant regions, and transition zones) and each region is processed independently with optimized filter characteristics. This segmentation allows the system to target noise reduction to specific problematic areas while preserving useful signal content in other regions, achieving more precise quality enhancement

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP1899962B1Audio codec post-filter
Publication Date: 2017.07.26 MICROSOFT TECHNOLOGY LICENSING LLC
  • EP1899962B1 patent drawingFigure 1
  • EP1899962B1 patent drawingFigure 2
  • EP1899962B1 patent drawingFigure 3

AI summary

Techniques and tools are described for processing reconstructed audio signals. For example, a reconstructed audio signal is filtered in the time domain using filter coefficients that are calculated, at least in part, in the frequency domain. As another example, producing a set of filter coefficients for filtering a reconstructed audio signal includes clipping one or more peaks of a set of coefficient values. As yet another example, for a sub-band codec, in a frequency region near an intersection between two sub-bands, a reconstructed composite signal is enhanced.