Asymmetric Overlap Windows for Low-Delay Audio Transform Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing transform audio coding techniques face challenges in reducing algorithmic delay while maintaining audio quality, particularly when a significant delay gain is sought, as asymmetrical windows with more zeros can introduce audible degradation due to high weighting values greater than 1, which is undesirable in digital signal processing.
Innovation Solution
A method employing asymmetrical weighting windows with four distinct portions: a rising interval, a constant interval of 1, a falling interval, and a constant interval of 0, optimized to reduce complexity and enhance audio quality by adjusting the durations of these intervals based on the desired delay, allowing for better frequency selectivity and reduced algorithmic delay without degrading audio quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If asymmetrical windows with more zeros are used to reduce algorithmic delay, then delay gain is improved, but audio quality deteriorates due to high weighting values greater than 1
Solution Approach 1:
The patent modifies the window function parameters by constraining the maximum weighting value to 1 and adjusting the distribution of zero samples. This changes the temporal characteristics of the window while maintaining frequency selectivity, thereby reducing algorithmic delay without introducing weighting values greater than 1 that would degrade audio quality
Solution Approach 2:
The patent employs asymmetrical window functions with a specific structure: a rising interval, a constant interval at value 1, a falling interval, and a constant interval at value 0. This asymmetrical design allows optimization of the transition regions while maintaining the constraint that weighting values do not exceed 1, thus achieving delay reduction without quality loss
2Loss of time
If symmetrical windows with zeros at start and end are used to reduce delay, then algorithmic delay is reduced, but frequency selectivity is lowered and audio quality deteriorates
Solution Approach 1:
The patent transitions from symmetrical window functions to asymmetrical ones, where the rising and falling intervals can have different characteristics. This asymmetry allows the window to maintain better frequency selectivity by optimizing the transition regions independently, while still achieving reduced algorithmic delay through the constant zero interval at the end
3Reliability
If asymmetrical windows with constrained maximum value of 1 are used, then audio quality is maintained, but the number of degrees of freedom for window design is reduced
Solution Approach 1:
The patent divides the window function into four distinct segments: a rising interval, a constant interval at value 1, a falling interval, and a constant interval at value 0. This segmentation allows each portion to be optimized independently for its specific function while maintaining the overall constraint that values do not exceed 1, thus preserving audio quality while providing structured design flexibility
Data Source
AI summary
Coding/decoding of a digital signal, consisting of successive blocks of samples, the coding being of the transform with overlap type and comprising, upon analysis, the application of a weighting window to two blocks of M successive samples. In particular, this weighting window is asymmetric and comprises four distinct portions extending successively over the two aforesaid blocks, with: a first portion, increasing over a first interval of samples, a second portion, constant at a value of 1 over a second interval, a third portion, decreasing over a third interval, and a fourth portion, constant at a value of 0 over a fourth interval.