Audio Peak Reduction Using Perceptual Loudness Modeling

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Audio playback systems face challenges in increasing perceived loudness without compromising sound quality, as traditional methods like dynamic range compression and all-pass filters often result in distortion and reduced loudness, especially at higher playback levels and for short audio signals.

Innovation Solution

A method that involves obtaining a digital audio signal, determining temporal amplitude peaks, and generating a new signal by reducing these peaks based on a predicted perceptual difference model to create a signal with a smaller crest factor, which is then amplified to increase perceived loudness while maintaining sound quality, using a signal processing device like a DSP.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Illumination intensity

If the volume is increased above a certain level, then the perceived loudness increases, but the reproduction accuracy deteriorates due to system limitations and driver over-excursion

Engineering Contradiction:
Improveperceived loudnessVSAvoidreproduction accuracy
Core Design Contradiction:
Illumination intensityVSManufacturing precision

Solution Approach 1:

The system performs preliminary peak detection and crest factor reduction on the audio signal before amplification. By pre-processing the signal to reduce temporal amplitude peaks and crest factor, the system prepares the signal to accommodate higher amplification levels without causing driver over-excursion or spectral imbalance, thus enabling increased perceived loudness while maintaining reproduction accuracy.

Inventive Principle:
Principle #10Preliminary action

2Illumination intensity

If dynamic range compression is applied to increase loudness, then the perceived loudness increases, but the dynamic range is reduced and audio quality deteriorates

Engineering Contradiction:
Improveperceived loudnessVSAvoiddynamic range
Core Design Contradiction:
Illumination intensityVSLoss of information

Solution Approach 1:

Instead of applying uniform dynamic range compression across the entire signal, the system applies localized crest factor reduction only to temporal amplitude peaks that exceed certain thresholds. This selective approach reduces the crest factor where needed while preserving the dynamic range and transient characteristics of the audio signal, thereby increasing perceived loudness without significantly degrading audio quality or losing dynamic information.

Inventive Principle:
Principle #3Local quality

3Manufacturing precision

If traditional peak reduction methods are used, then the temporal amplitude peaks are reduced, but distortion is introduced and loudness is reduced

Engineering Contradiction:
Improvefrequency response linearityVSAvoiddistortion
Core Design Contradiction:
Manufacturing precisionVSObject-generated harmful factors

Solution Approach 1:

The system employs dynamic thresholding and adaptive processing where the peak reduction amount is continuously adjusted based on the instantaneous signal characteristics. The crest factor reduction is applied adaptively - reducing peaks only when necessary and by the minimum amount required - rather than using fixed aggressive reduction. This dynamic approach minimizes distortion while achieving the desired frequency response linearity and maintaining adequate loudness.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP4138299A1A method for increasing perceived loudness of an audio data signal
Publication Date: 2023.02.22 BANG & OLUFSEN AS
  • EP4138299A1 patent drawingFigure 1~2
  • EP4138299A1 patent drawingFigure 3A~3B
  • EP4138299A1 patent drawingFigure 4~5

AI summary

Disclosed is a method for increasing a perceived loudness of an audio data signal comprising the steps of obtaining a first digital audio data signal; determining at least one temporal amplitude peak in the first digital audio data signal; generating a second digital audio data signal by reducing the at least one temporal amplitude peak in the first digital audio data signal based on a predicted perceptual difference model representing a predicted perceptual difference between the first digital audio data signal and a peak reduced version of the first digital audio data signal; and generating a third digital audio data signal by amplifying the second digital audio data signal so that a peak of the second digital audio data signal has a predetermined signal value, wherein a perceived loudness of the third digital audio data signal is larger than a perceived loudness of the first digital audio data signal. Further disclosed is a digital signal processing device, a loudspeaker device, a computer program product, and a digital audio data signal.