Frequency Domain Echo Cancellation via Transform Domain Processing

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Multi-channel communication systems face interference issues due to echo and near-end crosstalk, which degrade signal quality and require complex and power-intensive interference cancellation methods, especially at high sampling rates, leading to increased latency and cost in hardware implementation.

Innovation Solution

The method employs a frequency domain cancellation technique using Good-Thomas ordering, Winograd transforms, and Cooley-Tukey transforms to process signals in a transform domain, reducing interference by generating cancellation signals through weighting variables and inverse transforms, thereby minimizing latency and power consumption.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If time domain adaptive filters are used for interference cancellation, then interference cancellation accuracy is improved, but processing complexity and power consumption increase significantly

Engineering Contradiction:
Improveinterference cancellation accuracyVSAvoidprocessing complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent replaces time domain adaptive filtering with frequency domain processing using FFT transforms. This substitution changes the computational approach from iterative time domain adaptation to direct frequency domain multiplication, significantly reducing processing complexity while maintaining cancellation accuracy through the relationship: time domain convolution becomes frequency domain multiplication via FFT.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent transforms the signal from time domain to frequency domain parameters using FFT. By changing the domain of operation from time samples to frequency coefficients, the processing complexity is reduced because frequency domain multiplication is computationally simpler than time domain convolution, especially for long impulse responses.

Inventive Principle:
Principle #35Parameter changes

2Measurement precision

If time domain adaptive filters are used for interference cancellation, then interference cancellation accuracy is improved, but power consumption increases

Engineering Contradiction:
Improveinterference cancellation accuracyVSAvoidpower consumption
Core Design Contradiction:
Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The patent substitutes computationally intensive time domain adaptive filtering with more efficient frequency domain processing. The FFT-based approach reduces the number of multiplications required, directly lowering power consumption in hardware implementations while achieving the same interference cancellation through frequency domain multiplication instead of time domain convolution.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

By transforming to frequency domain parameters, the patent enables more efficient computation. The frequency domain representation allows cancellation filtering to be performed through simple multiplication of spectral coefficients, which consumes significantly less power than the iterative adaptation and convolution operations required in the time domain.

Inventive Principle:
Principle #35Parameter changes

3Productivity

If frequency domain canceller with complex Cooley-Tukey transforms is used, then processing efficiency is improved, but power consumption increases due to real adjust operations

Engineering Contradiction:
Improveprocessing efficiencyVSAvoidpower consumption
Core Design Contradiction:
ProductivityVSUse of energy by moving object

Solution Approach 1:

The patent segments the frequency domain processing into distinct stages: FFT transformation, frequency domain multiplication with transfer function, and inverse FFT transformation. This segmentation allows optimization of each stage independently, using efficient algorithms for each while minimizing overall power consumption, particularly by avoiding redundant real adjust operations through proper data organization.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent addresses the real vs. complex signal dimensionality issue by using frequency domain processing. By transforming real time-domain signals to complex frequency domain representations via FFT, the processing operates in a different dimensional space where mathematical operations are more efficient, and the inverse transform efficiently recovers the real time-domain result without requiring additional real adjust steps.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

4Device complexity

If block operations are used in frequency domain canceller, then processing complexity is reduced, but latency increases

Engineering Contradiction:
Improveprocessing complexityVSAvoidlatency
Core Design Contradiction:
Device complexityVSLoss of time

Solution Approach 1:

The patent implements continuous frequency domain processing by overlapping blocks of samples. Instead of processing completely separate blocks, the method maintains continuous operation by using overlapping windows and accumulating results, which reduces the effective latency while preserving the computational efficiency of block-based frequency domain processing.

Inventive Principle:
Principle #20Continuity of useful action

Data Source

PatentEP2255453B1Frequency domain echo and next cancellation
Publication Date: 2017.04.05 MARVELL INT LTD
  • EP2255453B1 patent drawingFigure 1
  • EP2255453B1 patent drawingFigure 2
  • EP2255453B1 patent drawingFigure 3

AI summary

A cancellation system is disclosed for processing incoming and outgoing signals in a transform domain to create a cancellation signal for reducing or removing unwanted interference. Data is ordered based on Good-Thomas indexing into a two dimensional array in a buffer. The two dimensional array may have lr rows and lw columns. From the buffer, the columns of data undergo a Winograd small transform. The rows of data undergo a Cooley-Tukey operation to complete the transform operation into the frequency domain. Multipliers scale the transformed data to generate a cancellation signal in the frequency domain. Inverse (Cooley-Tukey) and Winograd transforms perform inverse processing on the cancellation signal to return the cancellation signal or data to the time domain. Re-ordering the data and combination of the cancellation signal or data with incoming or outgoing signals achieve interference cancellation.