Adaptive Network Transforms Ambisonic Coefficients for Noise Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies face challenges in effectively removing interference from ambisonic signals when both the noise and audio signal are traveling in similar directions, leading to degraded audio quality.
Innovation Solution
The use of an adaptive network that applies constraints, such as directionality and signal type, to transform ambisonic coefficients, thereby spatially filtering out unwanted audio sources and improving signal-to-noise ratio.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional spatial filtering methods are used to remove interference from ambisonic signals, then the audio signal can be processed, but the signal-to-noise ratio is insufficient when noise and audio signal travel in similar directions
Solution Approach 1:
The patent transforms ambisonic coefficients from the spatial domain to the frequency domain using Fourier transform, enabling frequency-selective filtering. By operating in the frequency domain, the system can distinguish between desired audio signals and noise based on their spectral characteristics even when they arrive from similar directions, thereby improving signal-to-noise ratio without sacrificing interference removal effectiveness
Solution Approach 2:
The patent introduces an adaptive filter as an intermediary component that processes the frequency-transformed ambisonic coefficients. This adaptive filter learns optimal filtering parameters through training and dynamically adjusts to separate desired signals from noise, serving as a mediator that resolves the contradiction between maintaining signal integrity and removing interference
2Reliability
If adaptive networks with constraints are used to transform ambisonic coefficients, then the signal-to-noise ratio improves significantly, but the computational complexity increases
Solution Approach 1:
The patent performs preliminary Fourier transform of the ambisonic coefficients before applying adaptive filtering. This pre-processing step organizes the data in the frequency domain, making subsequent adaptive filtering more efficient and reducing the computational burden during real-time processing, thus achieving high signal-to-noise ratio improvement with manageable computational complexity
Solution Approach 2:
The patent segments the audio processing into distinct stages: Fourier transform stage, adaptive filtering stage, and inverse transform stage. Each segment can be optimized independently, allowing the system to achieve high signal-to-noise ratio through careful design of each segment while keeping overall computational complexity manageable through modular processing
Data Source
Figure 1
Figure 2A
Figure 2B
AI summary
A device includes a memory configured to store untransformed ambisonic coefficients at different time segments. The device also includes one or more processors configured to obtain the untransformed ambisonic coefficients at the different time segments, where the untransformed ambisonic coefficients at the different time segments represent a soundfield at the different time segments. The one or more processors are also configured to apply one adaptive network, based on a constraint, to the untransformed ambisonic coefficients at the different time segments to generate transformed ambisonic coefficients at the different time segments, wherein the transformed ambisonic coefficients at the different time segments represent a modified soundfield at the different time segments, that was modified based on the constraint.