Perceptual Noise Measurement in Audio Coders
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional audio compression techniques face high computational complexity in measuring perceptual noise due to the need for iterative excitation computation and bit rate adjustments, leading to poor audio quality at low bit rates.
Innovation Solution
A method that pre-computes noise-to-excitation ratio (NER) values by zeroing out spectral coefficients before the quantization loop, allowing for pre-computed NER values to be used for overall perceptual distortion calculation, reducing the need for iterative excitation computation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If iterative excitation computation is performed in the quality loop to maintain perceptual distortion criteria, then perceptual quality is improved, but computational complexity increases significantly
Solution Approach 1:
The patent pre-computes excitation values and NER (noise-to-excitation ratio) values before the quality loop begins. By calculating excitation for all critical bands in advance using the psychoacoustic model, the system eliminates the need for repeated excitation computations during iterative quality adjustments, thereby reducing computational complexity while maintaining perceptual quality.
Solution Approach 2:
The patent divides the computation into distinct phases: pre-computation of excitation and NER values for all critical bands, followed by separate quality loop iterations that use pre-computed values. This segmentation allows the heavy computational burden to be performed once during setup rather than repeatedly during optimization, reducing overall computational complexity.
2Reliability
If excitation is calculated each iteration in the quality loop to avoid zeroing out critical bands, then audio quality is maintained, but computational complexity increases
Solution Approach 1:
The patent performs the excitation calculation in advance before the quality loop starts. By pre-computing excitation values for all critical bands using the psychoacoustic model, the system avoids the need to recalculate excitation during each iteration of the quality loop, thereby maintaining audio quality while improving computational efficiency.
Solution Approach 2:
The patent creates a pre-computed copy of excitation values that can be reused during the quality loop iterations. Instead of recalculating excitation from scratch in each iteration, the system uses the pre-computed excitation values, significantly reducing the computational burden while maintaining the same audio quality assessment accuracy.
3Reliability
If conventional two-loop quantization is used to satisfy bit rate and perceptual distortion criteria, then audio quality is maintained, but processing time increases
Solution Approach 1:
The patent performs all necessary excitation and NER computations in advance before the quality loop begins. By pre-computing these values for all critical bands, the system eliminates repeated computations during iterative adjustments, thereby maintaining audio quality while significantly reducing processing time.
Solution Approach 2:
The patent ensures that useful computational actions are performed continuously during the pre-computation phase rather than being repeated during iterative loops. By calculating excitation and NER values once during setup and reusing them throughout the quality optimization process, the system maintains continuous progress toward quality goals without redundant computations.
Data Source
AI summary
A technique for computing perceptual noise in an audio signal that is computationally efficient. In one example embodiment, the technique includes computing perceptual noise in an input audio signal. The steps involve pre-computing NER (noise-to-excitation ratio) values associated with critical bands within a frame by zeroing out associated spectral coefficient values before the quantization loop, and also assuming bands with lower spectral energy than the band under consideration are zeroed out during quantization. When a critical band is zeroed out during quantization, the associated NER values which have been pre-computed are used in computing an overall perceptual distortion of the frame.