Integer MDCT Rounding Error Spectral Shaping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current audio coding methods, such as MPEG Layer3 and MPEG AAC, face challenges in achieving lossless coding due to accumulating rounding errors in integer transforms, particularly in the higher frequency range where audio signal energy is low, leading to inefficient coding and increased noise.
Innovation Solution
The development of an integer modified discrete cosine transform (IntMDCT) that uses lifting matrices and rounding steps to maintain integer output values, reducing rounding errors and enabling efficient lossless coding by shifting the rounding error to frequency ranges with high signal energy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If integer transforms with rounding steps are used to maintain integer output values, then coding efficiency for lossless compression is improved, but rounding errors accumulate particularly in higher frequency ranges leading to increased noise
Solution Approach 1:
The patent applies spectral shaping to modify the statistical properties of rounding errors, transforming them from white noise to colored noise with reduced energy in high-frequency ranges. This is achieved by applying a spectral shaping filter to the rounding errors generated during the integer MDCT transform, thereby reducing their harmful impact on audio quality while preserving coding efficiency
Solution Approach 2:
The patent converts the harmful rounding errors into a beneficial form by spectrally shaping them to match the psychoacoustic masking thresholds of the human ear. The rounding errors are deliberately shaped to have higher energy in frequency ranges where the human ear is less sensitive, thereby making them less perceptible while maintaining the benefits of integer arithmetic
2Ease of manufacture
If rounding errors are distributed uniformly across all frequency ranges, then implementation simplicity is maintained, but coding efficiency deteriorates in low-energy frequency ranges due to unnecessary noise
Solution Approach 1:
The patent modifies the spectral distribution parameter of rounding errors by applying spectral shaping filters. Instead of uniform distribution, the rounding errors are redistributed to concentrate energy in frequency ranges where it is least perceptible, thereby improving coding efficiency without significantly complicating the implementation through the use of efficient filter designs
Data Source
AI summary
When processing a signal having a sequence of discrete values, wherein there is a first frequency range, in which the signal has a high energy, and wherein there is a second frequency range, in which the signal has a low energy, the sequence of discrete values is first manipulated to obtain a sequence of manipulated values, so that at least one of the manipulated values is non-integer. Then the sequence of manipulated values is rounded to obtain a sequence of manipulated values. The rounding is formed to effect a spectral shaping of a generated rounding error so that a spectrally shaped rounding error has a higher energy in the first frequency range than in the second frequency range. By spectrally shaping the rounding error so that the rounding error does not have any energy either in the storage areas where there is no signal energy, an especially efficient coding is obtained particularly in connection with a lossless coding context.