Adaptive FIR Filter with Sparse Coefficient Memory for Echo Paths
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing FIR filter implementations in adaptive filtering systems, such as echo cancellation, require significant memory storage due to the need for storing all coefficients, including zero-valued ones, which can exceed available resources, especially in multi-channel applications.
Innovation Solution
Storing only non-zero valued coefficients with their corresponding memory addresses in a FIR filter, allowing for a variable number of coefficients per channel and eliminating the need for storing coefficients in order, thereby reducing memory usage while maintaining performance for demanding channels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all coefficients including zero-valued ones are stored in memory, then the filter can handle any echo path scenario, but the memory storage requirements become excessively large
Solution Approach 1:
The patent extracts and stores only the non-zero valued coefficients from the full set of filter coefficients. By identifying and removing the zero-valued coefficients that consume memory but contribute nothing to the filtering operation, the system achieves significant memory reduction while preserving the essential filtering capability for echo cancellation.
Solution Approach 2:
The patent implements a dynamic coefficient storage approach where the number of coefficients to be processed varies per channel based on the actual echo path characteristics. Each channel can have a different number of non-zero coefficients, allowing the system to adapt memory usage to the specific requirements of each channel rather than allocating fixed memory for all channels.
2Ease of manufacture
If a fixed number of coefficients are allocated per channel, then memory allocation is simplified, but channels with more demanding echo paths cannot be properly modeled
Solution Approach 1:
The patent makes the number of coefficients per channel dynamic rather than fixed. Each channel can have a different number of non-zero coefficients based on its specific echo path characteristics, allowing the system to adapt to varying channel requirements while maintaining manageable memory allocation through the use of compacted coefficient storage.
Solution Approach 2:
The patent changes the parameter of coefficient count from a fixed value to a variable that depends on the specific channel characteristics. By storing only non-zero coefficients and allowing the count to vary per channel, the system achieves both adaptability to different channel requirements and simplified memory management through compacted storage.
3Ease of operation
If coefficients are stored in sequential order with fixed positioning, then memory access is simplified, but the filter cannot efficiently handle channels with varying numbers of non-zero coefficients
Solution Approach 1:
The patent segments the coefficient storage into individual entries, where each entry contains a coefficient value and its corresponding position information. This segmentation allows each channel to have its own compacted list of non-zero coefficients with their positions, enabling efficient handling of varying coefficient counts while maintaining simple memory access through the structured entry format.
Solution Approach 2:
The patent adds a positional dimension to the coefficient storage by storing both the coefficient value and its original position index together. This dimensional change from storing only values to storing value-position pairs enables the system to handle variable numbers of coefficients per channel while maintaining the ability to reconstruct the original filter structure for efficient processing.
Data Source
AI summary
A FIR filter for use in an adaptive multi-channel filtering system, includes a first memory for storing data, and a second memory for storing filter coefficients. The second memory stores only non-zero valued coefficients or coefficients that are above a predetermined magnitude threshold such that the overall number of coefficients processed is significantly reduced.