Ambisonic Noise Reduction by Order to Preserve Direction Cues
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional noise reduction techniques for microphone arrays used in ambisonic signal capture impair audio quality and introduce errors in direction information, and applying noise reduction at loudspeakers during playback can cause audio quality artefacts.
Innovation Solution
A device and method that apply distinct gain factors to ambisonic signals of different orders based on noise data, performing noise reduction independently for each order to reduce noise while preserving directional information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If conventional noise reduction techniques (Wiener filtering and spectral subtraction) are applied at the microphones, then noise is reduced, but audio quality is impaired and errors in direction information are introduced
Solution Approach 1:
The noise reduction process is segmented by applying different gain factors to ambisonic signals of different orders (first order, second order, etc.) based on their respective noise characteristics. This allows targeted noise reduction for each order while preserving the directional information unique to each order, resolving the contradiction between noise reduction and directional accuracy.
Solution Approach 2:
Different noise reduction strategies (distinct gain factors) are applied to different parts of the ambisonic signal (different orders) according to their local noise characteristics. This localised approach ensures that noise is reduced in each order without uniformly degrading the audio quality and directional information across all orders.
2Object-affected harmful factors
If Wiener filtering and spectral subtraction are applied independently at loudspeakers during playback, then noise is reduced, but audio quality artefacts are introduced when loudspeaker contributions are added
Solution Approach 1:
Noise reduction is performed in advance during the ambisonic signal capture and processing stage, rather than during playback at the loudspeakers. By applying order-specific noise reduction to the ambisonic coefficients before rendering to loudspeakers, the system eliminates noise without introducing playback artefacts, as the noise reduction is baked into the signal representation itself.
3Device complexity
If a single noise reduction operation is applied to all ambisonic signals, then processing is simplified, but directional information errors are introduced
Solution Approach 1:
The ambisonic signal processing is segmented into multiple independent noise reduction operations, one for each ambisonic order. Each order receives a tailored gain factor based on its specific noise characteristics, preserving the directional information unique to each order while maintaining manageable processing complexity through modular operation.
Solution Approach 2:
The noise reduction approach changes the parameter of gain application from uniform across all orders to order-specific variable gains. By adjusting the gain factor parameter independently for each ambisonic order based on measured noise characteristics, the system achieves both directional accuracy and controlled complexity.
Data Source
AI summary
A device to apply noise reduction to ambisonic signals includes a memory configured to store noise data corresponding to microphones in a microphone array. A processor is configured to perform signal processing operations on signals captured by microphones in the microphone array to generate multiple sets of ambisonic signals including a first set corresponding to a first particular ambisonic order and a second set corresponding to a second particular ambisonic order. The processor is configured to perform a first noise reduction operation that includes applying a first gain factor to each ambisonic signal in the first set and to perform a second noise reduction operation that includes applying a second gain factor to each ambisonic signal in the second set. The first gain factor and the second gain factor are based on the noise data, and the second gain factor is distinct from the first gain factor.