Audio Frame Decoding with LPC Smoothing for Click-Free Transitions

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing audio signal processing methods using long-term prediction filters often introduce discontinuities between filtered frames, leading to artifacts like 'clicks' due to mismatched filter parameters, and prior solutions can cause additional distortion, especially in harmonic signals with fast-changing pitch.

Innovation Solution

The method employs linear predictive filtering to remove discontinuities by modifying the beginning portion of the filtered current frame using initial states from the past frame, avoiding the use of past frame filter parameters and utilizing Levinson-Durbin algorithm for LPC filter estimation to mask spectral shape and reduce distortion.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Object-affected harmful factors

If long-term prediction filter is used to attenuate harmonics in audio signal, then harmonic suppression is improved, but discontinuities are introduced between filtered frames

Engineering Contradiction:
Improveharmonic suppressionVSAvoiddiscontinuities between frames
Core Design Contradiction:
Object-affected harmful factorsVSObject-generated harmful factors

Solution Approach 1:

An overlap region is introduced as an intermediary between consecutive filtered frames. This overlap region contains both the end portion of the previous frame and the beginning portion of the current frame, allowing smooth transition and eliminating discontinuities while maintaining harmonic suppression benefits

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The overlap region is prepared in advance by including it in both the previous and current frame processing. This preliminary action ensures that when frames are concatenated, the transition is already smoothed out, preventing discontinuities before they occur

Inventive Principle:
Principle #10Preliminary action

2Object-generated harmful factors

If overlap-add operation is performed to remove discontinuities, then discontinuity removal is improved, but additional distortion is introduced in harmonic signals with fast-changing pitch

Engineering Contradiction:
Improvediscontinuity removalVSAvoidadditional distortion
Core Design Contradiction:
Object-generated harmful factorsVSObject-affected harmful factors

Solution Approach 1:

Different processing approaches are applied to different regions of the audio signal. The overlap region receives special handling with linear predictive coding to match spectral characteristics, while the rest of the frames are processed normally. This localized approach removes discontinuities without introducing distortion in regions where it matters most

Inventive Principle:
Principle #3Local quality

3Device complexity

If filter parameters from past frame are used for current frame, then processing simplicity is improved, but spectral shape mismatch causes distortion

Engineering Contradiction:
Improveprocessing simplicityVSAvoidspectral shape mismatch distortion
Core Design Contradiction:
Device complexityVSObject-affected harmful factors

Solution Approach 1:

Filter parameters are dynamically changed for the overlap region based on the spectral characteristics of the current frame, rather than using fixed parameters from the past frame. This allows the filter to adapt to spectral shape changes, eliminating mismatch distortion while maintaining processing efficiency

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3407351B1Methods for encoding and decoding an audio signal, audio decoder and audio encoder
Publication Date: 2020.01.01 FRAUNHOFER GESELLSCHAFT ZUR FORDERUNG DER ANGEWANDTEN FORSCHUNG EV
  • EP3407351B1 patent drawingFigure 1~2
  • EP3407351B1 patent drawingFigure 3
  • EP3407351B1 patent drawingFigure 4

AI summary

A method is described that processes an audio signal (100). A discontinuity between a filtered past frame and a filtered current frame of the audio signal is removed using linear predictive filtering (102, 110, 112).