Low-Delay Synthesis Filterbank Using Extended Windowing

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Digital audio systems face challenges in balancing bitrate, computational complexity, quality, and delay, often requiring compromises in these parameters, especially in applications requiring low delay without reducing audio quality.

Innovation Solution

The implementation of a synthesis filterbank with a frequency/time converter, windower, and overlap/adder, utilizing a low-delay window function that extends windowing into the past and incorporates multiple overlaps to reduce delay while maintaining quality, allowing for efficient processing and encoding/decoding.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of time

If traditional windowing with 20 ms delay is used, then audio quality is maintained, but processing delay increases

Engineering Contradiction:
Improveprocessing delayVSAvoidaudio quality
Core Design Contradiction:
Loss of timeVSReliability

Solution Approach 1:

The patent applies preliminary action by extending the window function into the past (using future samples in the window calculation) to achieve low-delay processing. This allows the system to prepare and process audio data with reduced latency while maintaining quality through careful design of the window function and overlap-add parameters.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent changes key parameters including the window function definition (extending into past/future), overlap factor (using at least 3 overlapping frames), and sample advance value to achieve the desired delay reduction from 20 ms to 15 ms while maintaining audio quality through optimized parameter combinations.

Inventive Principle:
Principle #35Parameter changes

2Loss of time

If multiple overlaps are incorporated to reduce delay, then processing delay decreases, but device complexity increases

Engineering Contradiction:
Improveprocessing delayVSAvoidfilterbank complexity
Core Design Contradiction:
Loss of timeVSDevice complexity

Solution Approach 1:

The patent applies segmentation by dividing the audio signal into multiple overlapping frames with at least 3 segments per processing interval. This segmentation approach enables reduced delay through multiple overlaps while managing complexity through systematic organization of the filterbank structure and processing stages.

Inventive Principle:
Principle #1Segmentation

3Loss of time

If low-delay window function is used, then processing delay reduces, but computational complexity increases

Engineering Contradiction:
Improveprocessing delayVSAvoidcomputational complexity
Core Design Contradiction:
Loss of timeVSDevice complexity

Solution Approach 1:

The patent applies partial or excessive action by using a window function that extends beyond the current frame boundaries into past and future samples. This excessive windowing approach achieves low delay (15 ms) but requires additional computational resources for handling the extended window samples and multiple overlaps.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP2884490B1Analysis filterbank, synthesis filterbank, encoder, decoder, mixer and conferencing system
Publication Date: 2016.06.29 FRAUNHOFER GESELLSCHAFT ZUR FORDERUNG DER ANGEWANDTEN FORSCHUNG EV
  • EP2884490B1 patent drawingFigure 1
  • EP2884490B1 patent drawingFigure 2
  • EP2884490B1 patent drawingFigure 3

AI summary

An embodiment of an analysis filterbank for filtering a plurality of time domain input frames, wherein an input frame comprises a number of ordered input samples, comprises a windower configured to generating a plurality of windowed frames, wherein a windowed frame comprises a plurality of windowed samples, wherein the windower is configured to process the plurality of input frames in an overlapping manner using a sample advance value, wherein the sample advance value is less than the number of ordered input samples of an input frame divided by two, and a time/frequency converter configured to providing an output frame comprising a number of output values, wherein an output frame is a spectral representation of a windowed frame.