Audio Data Regularization via Periodic Phase Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing audio data modification methods lack the ability to significantly improve the quality and speed of audio data processing, particularly in achieving realistic and natural audio signals, as they often rely on iterative processes that do not effectively address phase data modification.
Innovation Solution
The method involves transforming audio data into a time-frequency domain, iteratively modifying magnitude and phase data, and performing specific operations every fifth loop pass, including transforming complex data into more real data and combining it with initial data to enhance audio quality, using techniques such as regularization and phase optimization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If iterative processes are used to modify audio data, then audio data can be modified step-by-step, but the quality and speed of audio data processing cannot be significantly improved
Solution Approach 1:
The patent applies periodic action by performing specific operations at regular intervals during the iterative process - specifically every fifth loop pass. This periodic intervention with projected regularization and phase optimization maintains audio quality while preventing excessive processing time, resolving the contradiction between manufacturing precision and productivity.
Solution Approach 2:
The patent changes parameters by transforming audio data between time domain and frequency-time domain, and by modifying magnitude and phase data differently at different iteration intervals. This parameter transformation enables efficient processing while maintaining high audio quality, addressing both speed and quality requirements.
2Manufacturing precision
If phase data is modified in every loop pass, then audio quality may improve, but processing time increases significantly
Solution Approach 1:
The patent implements periodic action by performing full phase data modification and projected regularization only every fifth loop pass, rather than in every iteration. This periodic approach maintains audio quality through sufficient phase optimization while dramatically reducing the cumulative processing time compared to continuous phase modification.
Solution Approach 2:
The patent applies partial action by performing projected regularization selectively at specific intervals (every fifth pass) rather than continuously. This partial application of phase modification is sufficient to achieve the desired audio quality without the excessive time cost of applying it in every iteration.
3Manufacturing precision
If projected regularization is applied at every loop pass, then audio quality improves, but computational complexity increases
Solution Approach 1:
The patent reduces computational complexity by applying projected regularization periodically every fifth loop pass rather than at every iteration. This periodic application maintains the audio quality benefits of projected regularization while significantly reducing the overall computational burden and complexity of the processing system.
4Manufacturing precision
If iterative modification steps are applied multiple times, then audio reconstruction quality improves, but the process becomes more complex
Solution Approach 1:
The patent manages process complexity by implementing a structured periodic pattern where specific modification steps are applied every fifth loop pass. This regular periodic structure provides a predictable and manageable complexity level while still achieving high audio reconstruction quality through multiple iterations of magnitude and phase modification.
Data Source
Figure 1~3
Figure 4~6
Figure 7
AI summary
A method of modification of audio data is provided, to improve the quality of the audio modification or reconstruction, or to improve the speed of such reconstruction or modification and produce more realistic audio data. Realistic audio data is audio data that is generated in natural events like talking, or singing, or a vehicle passing by, and is not generated only by artificially constructing audio data, like in a synthesizer. This will lead to audio data that will be perceived more likely as natural or unmodified audio signal when being played back to human beings. The method involves modification of some part of transformed audio data, especially phase data.