Audio Similarity Detection via Spectral SAD Analysis

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing methods for identifying similarities between original and candidate audio files are not robust enough to handle variations such as different time scales, processing techniques, and modifications, leading to inaccurate matches.

Innovation Solution

A method and system that utilize a novel Sum of Absolute Difference (SAD)-based 'Signature Score' measure, combined with time-frequency mapping, linear interpolation, and histogram analysis, to accurately match audio files despite variations in processing and performance.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of manufacture

If block correlation in time domain is used for detecting matches, then the technique is simple to implement, but it fails when samples have different time scales or are highly processed

Engineering Contradiction:

Improveease of implementationVSAvoidmatching accuracy

Core Design Contradiction:

Ease of manufactureVSReliability

Solution Approach 1:

The patent replaces the mechanical time-domain block correlation method with a spectral-based approach using Fast Fourier Transform (FFT). This substitution allows the system to compare audio files in the frequency domain, making it invariant to time-scale modifications and robust against heavy processing while maintaining computational efficiency.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent transforms the audio comparison from time-domain parameters to frequency-domain parameters by applying FFT. This parameter change enables the system to detect similarities based on spectral characteristics rather than temporal alignment, resolving the issue of time-scale differences and processing variations.

Inventive Principle:

Principle #35Parameter changes

2Device complexity

If simple technique is used for estimating time scale difference using end point of samples, then the calculation is straightforward, but it fails due to truncation and silence additions in processed samples

Engineering Contradiction:

Improvecalculation complexityVSAvoidtime scale estimation accuracy

Core Design Contradiction:

Device complexityVSMeasurement precision

Solution Approach 1:

The patent replaces the endpoint-based time scale estimation with a spectral ratio method. By computing the ratio of spectral energies at different time points, the system can accurately estimate time scale factors even when samples are truncated or have silence added, as spectral characteristics remain preserved.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Reliability

If machine learning methods such as deep learning algorithms are applied, then the matching can handle complex variations, but training the model is extremely difficult requiring large training database

Engineering Contradiction:

Improvematching robustnessVSAvoidmodel training complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent employs a lightweight spectral comparison method that does not require complex machine learning models or large training databases. By using FFT-based spectral analysis and simple energy ratio calculations, the system achieves robust matching while avoiding the computational burden and data requirements of deep learning approaches.

Inventive Principle:

Principle #27Cheap short-living objects (Disposable)

Solution Approach 2:

The patent substitutes complex machine learning systems with a deterministic spectral analysis approach. This replacement maintains matching robustness through mathematical properties of the Fourier transform while eliminating the need for model training and large datasets.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

4Adaptability or versatility

If the audio files are heavily processed with different codecs, sample rates, and bit-depths, then the adaptability to real-world variations is improved, but the accuracy of similarity detection deteriorates

Engineering Contradiction:

Improveprocessing variation toleranceVSAvoidsimilarity detection accuracy

Core Design Contradiction:

Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent uses spectral analysis in the frequency domain which is inherently more robust to processing variations than time-domain methods. By comparing spectral energies and patterns rather than raw waveforms, the system maintains high accuracy even when files are encoded with different codecs, sample rates, or bit-depths.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent transforms audio files into the frequency domain using FFT, changing the representation parameters from time-domain samples to frequency-domain spectral coefficients. This parameter transformation makes the comparison invariant to many processing operations that preserve spectral characteristics.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS20250069618A1Method and System for Identifying Similarity Between Two Audio Tracks

Publication Date: 2025.02.27 AUDIO TECH & CODECS

AI summary

The invention provides a method for identifying similarity between two audio files or tracks. The method comprises receiving a processed audio file and an original audio file, uncompressing the processed audio file, applying global loudness normalization and short-term loudness normalization on the processed audio file and the original audio file, converting the processed audio file and the original audio file into processed spectral image by time-frequency mapping, scaling, using linear interpolation, the processed spectral image, dividing the scaled-up processed spectral image into slices, searching for minimum Sum of Absolute Difference (SAD), using original spectral image as reference, for each slice.