Neural Audio Equalization for Perceived Spectrum Balancing

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current digital audio workstations (DAWs) lack an efficient computer-implemented aid for automatic profile equalization in audio production, particularly in balancing the perceived spectrum of audio content during recording, mixing, and mastering processes.

Innovation Solution

A method utilizing a neural network to determine model parameters from an input audio window, which automatically equalizes the audio by matching its spectrum to a target spectrum based on psychoacoustically weighted contours, such as ISO226-2003 equal-loudness-level contours, to optimize the perceived loudness and balance the audio content.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual spectrum balancing is performed during mastering, then audio quality and perceived spectrum balance are improved, but time consumption and operational complexity increase

Engineering Contradiction:

Improvespectrum balance accuracyVSAvoidmastering time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system performs automatic spectrum balancing without requiring manual intervention. The neural network automatically analyzes the input audio, determines the target spectrum based on monitoring loudness, and applies equalization parameters to achieve balanced output, allowing the system to serve itself in the mastering process.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The patent replaces manual auditory analysis and adjustment with an automated neural network system. The neural network substitutes human experts' mechanical process of listening and adjusting equalization parameters with an automated computational system that processes audio spectra and applies corrections based on learned patterns from training data.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Productivity

If automatic equalization is implemented without neural network training, then processing speed is improved, but equalization accuracy and perceived spectrum balance deteriorate

Engineering Contradiction:

Improveprocessing speedVSAvoidspectrum balance accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The neural network is trained in advance on a large dataset of audio files with manually adjusted equalization examples. This preliminary training phase stores learned patterns and relationships between input spectra and target spectra, enabling the network to perform accurate automatic equalization during actual processing without requiring real-time manual intervention.

Inventive Principle:

Principle #10Preliminary action

3Measurement precision

If complex neural network models are used for spectrum analysis, then equalization accuracy is improved, but device complexity and computational resources increase

Engineering Contradiction:

Improvespectrum balance accuracyVSAvoidsystem complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The system transforms the audio processing problem into a parameter optimization problem. The neural network learns to predict optimal equalization parameters (gain adjustments at different frequencies) based on input audio characteristics. By changing the approach from complex real-time spectral manipulation to parameter prediction, the system achieves accurate results with a manageable computational model.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentEP3688756B1Method and electronic device

Publication Date: 2022.11.09 SONY EUROPE BV

Figure 1
Figure 2
Figure 3

AI summary

A method comprising determining feature values of an input audio window and determining model parameters for the input audio window based on processing of feature values using a neural network.