Adaptive DSHT Compression of HOA Audio to Reduce Noise Unmasking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Higher Order Ambisonics (HOA) audio signals experience increased coding noise and noise unmasking due to high cross-correlation between channels, especially after matrixing operations and transformation to the spatial domain for compression, which existing perceptual coders fail to address effectively.
Innovation Solution
Implement an adaptive Discrete Spherical Harmonics Transform (aDSHT) that decorrelates HOA channels based on their spatial properties, minimizing noise unmasking by rotating the sampling grid and integrating it within a compressive coder architecture, reducing the need for transmitting extensive side information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If independent perceptual coders are applied to individual Ambisonics coefficient channels, then compression is achieved, but coding noise increases due to noise unmasking after matrixing
Solution Approach 1:
The patent applies a preliminary decorrelation transformation (Karhunen-Loève transform or discrete spherical harmonics transform) to the Ambisonics coefficient channels before compression. This preliminary action reorganizes the channel correlations so that when independent perceptual coding is applied, the resulting coding noise does not cause unmasking artifacts after the subsequent matrixing operation.
2Adaptability or versatility
If HOA signals are transformed to spatial domain by Discrete Spherical Harmonics Transform prior to compression, then spatial processing is enabled, but noise unmasking occurs during matrixing
Solution Approach 1:
The patent performs a preliminary transformation to a decorrelated domain (either through Karhunen-Loève transform or discrete spherical harmonics transform) before applying independent perceptual coding. This ensures that when the signals are later transformed to spatial domain and matrixed, the coding noise does not cause unmasking artifacts.
Solution Approach 2:
The patent changes the domain parameters by transforming from the original Ambisonics coefficient domain to a decorrelated domain where channels have reduced cross-correlation. This parameter change in the transformation domain allows for effective independent compression while avoiding noise unmasking in the spatial domain.
3Quantity of substance
If high cross-correlation between channels is present, then compact representation is maintained, but coding noise increases after matrixing
Solution Approach 1:
The patent applies a preliminary decorrelation transformation that reorganizes the highly correlated channels into a domain with reduced cross-correlation. This preliminary action preserves the information content while changing the correlation structure, enabling effective independent compression without noise unmasking artifacts.
Data Source
Figure 1~3
Figure 4a~4b
Figure 5a~6
AI summary
A method for encoding multi-channel HOA audio signals for noise reduction comprises steps of decorrelating (81) the channels using an inverse adaptive DSHT, the inverse adaptive DSHT comprising a rotation operation (330) and an inverse DSHT (810), with the rotation operation rotating the spatial sampling grid of the iDSHT, perceptually encoding (82) each of the decorrelated channels, encoding rotation information (SI), the rotation information comprising parameters defining said rotation operation, and transmitting or storing the perceptually encoded audio channels and the encoded rotation information.