Partitioned Frequency-Domain Adaptive Filter With Corrected Gradient Updates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Partitioned block frequency domain adaptive filter devices face high complexity, which affects their performance and efficiency in applications like acoustic echo cancellation and noise reduction.
Innovation Solution
A partitioned block frequency domain adaptive filter device with a frequency domain adaptive filter and parallel filter update blocks, using an approximated constraining matrix to reduce complexity while maintaining convergence rate, and a correction module to adjust filter coefficients based on cumulative errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a full constraining matrix is used in partitioned block frequency domain adaptive filter devices, then filtering accuracy is improved, but device complexity increases significantly
Solution Approach 1:
The filter coefficients are divided into multiple blocks, and the constraining matrix operations are segmented to process only relevant blocks. This segmentation allows the system to maintain filtering accuracy for critical frequency regions while reducing overall computational complexity by avoiding unnecessary matrix operations on all blocks.
Solution Approach 2:
The patent applies the full constraining matrix selectively to specific frequency blocks that require high precision filtering, while using simplified or approximate constraints for other blocks. This local quality approach ensures that filtering accuracy is maintained where needed most, while reducing device complexity in less critical regions.
2Productivity
If parallel filter update blocks are used to reduce processing time, then productivity is improved, but device complexity increases
Solution Approach 1:
The filter update process is segmented into multiple parallel blocks that can be processed simultaneously. Each block handles a specific portion of the frequency spectrum, enabling concurrent processing that improves productivity while keeping individual block complexity manageable.
Solution Approach 2:
The patent implements partial parallelization where only the most computationally intensive filter update blocks are executed in parallel, while less demanding blocks are processed sequentially. This selective parallel approach improves processing speed for critical operations without proportionally increasing device complexity.
3Device complexity
If an approximated constraining matrix is used to reduce complexity, then device complexity is reduced, but convergence rate may be impacted
Solution Approach 1:
The patent applies approximated constraining matrices selectively to frequency blocks where high precision is not critical, while using full constraining matrices for blocks that require accurate convergence. This local differentiation maintains overall convergence rate while reducing device complexity in appropriate regions.
Solution Approach 2:
The patent dynamically adjusts the approximation level of the constraining matrix based on operational conditions, signal characteristics, and convergence requirements. By changing the parameter of matrix approximation dynamically, the system maintains convergence rate when needed while reducing complexity during stable operation or in less critical frequency regions.
Data Source
AI summary
A partitioned block frequency domain adaptive filter device includes a frequency domain adaptive filter configured for filtering a frequency domain representation of a time domain input signal depending on a set of filter coefficients consisting of a plurality of blocks of filter coefficients in order to produce a filtered signal; a plurality of parallel arranged filter update blocks; wherein each of the filter update blocks includes an adaptation module configured for executing an adaptation sequence including the steps of calculating an approximation of a constrained gradient update for the filter coefficients of the respective block of filter coefficients, and calculating a cumulative error introduced on the unconstrained gradient update; wherein each of the filter update blocks includes a correction module configured for executing a correction sequence including the steps of calculating a corrected constrained gradient update for the filter coefficients of the respective block of filter coefficients.