Single-Track Audio Source Separation With Iterative Artifact Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing audio source separation techniques are not optimized to generate high-quality audio stems from low-quality, single-track, noisy sound mixtures, particularly from older recordings, which are common in the music and film industries.
Innovation Solution
A system and method using machine learning models, including neural networks, to separate and enhance audio sources from single-track recordings, employing self-iterative training, windowing functions, and post-processing to remove artifacts, with user-guided fine-tuning for improved fidelity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If existing audio source separation techniques are used, then the process can be completed, but the output audio stems are not high-quality enough for professional use
Solution Approach 1:
The system implements self-iterative training where the audio separation model processes output stems through a quality assessment module, and the results feed back into retraining the model. This feedback loop continuously improves the model's ability to produce high-quality, high-fidelity audio stems from low-quality recordings.
Solution Approach 2:
The system performs preliminary actions by training the audio separation model on synthetic data that simulates the characteristics of low-quality historical recordings before processing actual audio. This pre-training on representative data prepares the model to handle the specific challenges of degraded audio signals.
2Manufacturing precision
If machine learning models are trained iteratively, then audio quality improves, but processing time increases
Solution Approach 1:
The system performs preliminary training on synthetic data and pre-processes audio signals before they enter the main iterative training loop. This preliminary action reduces the computational burden during the main processing phase, allowing quality improvement without proportional time increases.
Solution Approach 2:
The system applies partial iterative refinement rather than complete retraining from scratch. By updating the model incrementally based on quality assessments of intermediate outputs, the system achieves significant quality improvements with reduced computational time compared to full retraining.
3Adaptability or versatility
If the audio separation model processes complex mixtures, then more audio sources can be separated, but the complexity of processing increases
Solution Approach 1:
The system segments the audio separation task into multiple specialized processing stages, with each stage handling specific aspects of source separation. This segmentation allows the model to handle complex mixtures of multiple audio sources by breaking down the processing into manageable, specialized operations.
Solution Approach 2:
The processing pipeline is divided into sequential stages where each stage processes specific aspects of the audio mixture. This segmentation reduces overall processing complexity by distributing the computational burden across multiple simpler, specialized operations rather than one complex monolithic process.
Data Source
AI summary
Systems and methods includes receiving a single-track audio input stream having a mixture of audio signals generated from a plurality of sources, training an audio source separation model using, at least in part, the received single-track audio input stream, and separating audio sources, using the audio source separation model, from the audio input stream in accordance with one or more processing recipes to generate a plurality of source separated output stems. The audio separation model is trained to receive the single-track audio input stream and generate a plurality of audio stems corresponding to one or more audio sources of the plurality of sources.