Adaptive Noise Filter Coefficient Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speech processing systems face high computational and memory costs, along with poor adaptation to sampling rates, when reducing noise signals using techniques like spectral subtraction and neural networks.
Innovation Solution
An adaptive filter system with two stages, using a learning rule to continuously reduce coefficient values, implemented with a FIR filter and a second filter for sustained noise reduction, significantly reduces computational and memory requirements while allowing for efficient adaptation to different sampling rates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If spectral subtraction is used to reduce noise signals, then noise reduction is achieved, but computational cost and memory requirement increase significantly
Solution Approach 1:
The patent extracts only the essential information needed for noise reduction by using a simplified filter model that processes only the most relevant signal components. Instead of processing the complete spectral representation, the system uses a reduced-order model that captures the fundamental noise characteristics, thereby lowering computational requirements while maintaining effective noise suppression.
Solution Approach 2:
The patent employs a simplified filter structure with fewer coefficients that can be quickly updated and discarded. The filter uses a reduced number of parameters that are easier and faster to compute, replacing the complex spectral subtraction method with a more efficient algorithm that achieves comparable noise reduction with lower computational cost.
2Object-affected harmful factors
If spectral subtraction is used to reduce noise signals, then noise reduction is achieved, but memory requirement increases significantly
Solution Approach 1:
The patent extracts only the essential information needed for noise reduction by using a simplified filter model that processes only the most relevant signal components. Instead of processing the complete spectral representation, the system uses a reduced-order model that captures the fundamental noise characteristics, thereby lowering computational requirements while maintaining effective noise suppression.
3Object-affected harmful factors
If spectral subtraction parameters are adapted, then noise reduction is improved, but adaptation to other sampling rates becomes difficult
Solution Approach 1:
The patent designs a filter model that is universally applicable across different sampling rates. The simplified filter structure uses normalized coefficients and a general form that can be adapted to various sampling frequencies without requiring complete parameter reconfiguration. This universal design allows the same filter architecture to function effectively at different sampling rates, improving adaptability while maintaining noise reduction performance.
4Object-affected harmful factors
If complex filtering techniques are used to reduce noise, then noise reduction is achieved, but processing time increases
Solution Approach 1:
The patent extracts only the essential information needed for noise reduction by using a simplified filter model that processes only the most relevant signal components. Instead of processing the complete spectral representation, the system uses a reduced-order model that captures the fundamental noise characteristics, thereby lowering computational requirements while maintaining effective noise suppression.
Solution Approach 2:
The patent employs a simplified filter structure with fewer coefficients that can be quickly updated and discarded. The filter uses a reduced number of parameters that are easier and faster to compute, replacing the complex spectral subtraction method with a more efficient algorithm that achieves comparable noise reduction with lower computational cost.
Data Source
AI summary
An audio input signal is filtered using an adaptive filter to generate a prediction output signal with reduced noise, wherein the filter is implemented using a plurality of coefficients to generate a plurality of prediction errors and to generate an error from the plurality of prediction errors, wherein the absolute values of the coefficients are continuously reduced by a plurality of reduction parameters.