Noise Injection Algorithm for Audio Coding Sparsity Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing audio coding schemes face challenges in maintaining high perceptual quality while minimizing information transmission, particularly in handling non-noise-like signals and ensuring efficient coding of audio signals with significant portions of zero energy in the frequency domain, which can result in artifacts like dullness and lack of naturalness.
Innovation Solution
A noise injection algorithm is implemented to adjust gain and spectral shape of injected noise, using a sparsity factor to control noise level and modulate the noise injection gain based on the sparsity of the audio signal's energy distribution, ensuring that noise is only added where necessary, such as in noise-like regions, and shaping the noise spectrum to match the coded signal.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If noise is injected into zero-energy regions of the coded spectrum, then perceptual quality is improved by reducing artifacts, but information transmission efficiency deteriorates due to additional noise data
Solution Approach 1:
The patent applies noise injection selectively only to zero-energy regions of the coded spectrum where it is perceptually beneficial, rather than uniformly across the entire spectrum. The system identifies regions with significant zero energy and applies noise injection only there, maintaining efficiency in non-zero regions while improving perceptual quality in appropriate regions.
Solution Approach 2:
The patent dynamically adjusts the noise injection level based on the sparsity factor, which measures the proportion of zero-energy coefficients. When sparsity is high (indicating many zero coefficients), noise injection is applied more aggressively. When sparsity is low, noise injection is reduced or eliminated, optimizing the trade-off between quality and transmission efficiency.
2Manufacturing precision
If noise injection level is increased to improve perceptual quality in noise-like regions, then naturalness is enhanced, but coding efficiency deteriorates due to increased information transmission
Solution Approach 1:
The patent applies noise injection partially rather than fully across the spectrum. By using the sparsity factor to determine the degree of noise injection, the system applies just enough noise to improve perceptual quality in noise-like regions without over-injecting, which would waste transmission bits and reduce coding efficiency.
3Productivity
If sparsity factor is used to modulate noise injection gain, then noise is added only where necessary improving efficiency, but device complexity increases due to additional processing
Solution Approach 1:
The sparsity factor is calculated directly from the coded spectrum coefficients that are already available in the system, without requiring additional complex analysis or external data. The existing coefficient information serves dual purposes: both for decoding and for determining noise injection levels, eliminating the need for separate complexity-intensive processing stages.
Data Source
Figure 1
Figure 2
Figure 3A~3B
AI summary
A scheme for injecting noise at uncoded elements of a spectrum is controlled according to a measure of a distribution of energy of the original spectrum among the locations of the uncoded elements.