Kernel Transformation for Echo Cancellation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Wireless devices face interference due to signal leakage, or echo, between their transmit and receive components, which affects the quality of received signals, especially when operating with multiple wireless protocols like LTE and Wi-Fi, leading to non-linear distortion and interference.
Innovation Solution
The implementation of a kernel transformation system (KTS) that uses whitening kernels through Cholesky decomposition or Gram-Schmidt orthogonalization to regenerate non-linear terms in the digital domain, reducing Peak Average Power Ratio (PAPR) and minimizing echo cancellation errors by generating orthogonal kernels, which are then used to cancel out echo signals in the receive chain.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional echo cancellation methods are used, then echo interference can be reduced, but non-linear distortion causes convergence instability and high Mean Squared Error (MSE)
Solution Approach 1:
The patent transforms the kernel parameters from correlated to orthogonal by applying a transformation matrix to the autocorrelation matrix. This parameter transformation changes the statistical properties of the kernel, making them orthogonal and thereby stabilizing adaptive filter convergence while reducing MSE in the presence of non-linear distortion
Solution Approach 2:
The patent introduces an intermediate transformation step that processes the original kernel through a transformation matrix derived from the autocorrelation matrix. This intermediary transformation creates orthogonal kernels that serve as a bridge between the original correlated kernels and the final echo cancellation output, improving convergence stability
2Measurement precision
If high-order kernels are used to improve echo cancellation, then cancellation accuracy increases, but Peak Average Power Ratio (PAPR) increases causing instability
Solution Approach 1:
The patent applies a parameter transformation to the kernel that changes its statistical properties. By transforming correlated high-order kernels into orthogonal kernels through the transformation matrix, the patent maintains the high cancellation accuracy of high-order kernels while reducing the PAPR-induced instability that would otherwise prevent convergence
3Ease of manufacture
If correlated kernels are used for echo cancellation, then implementation is simpler, but convergence stability deteriorates due to non-linear distortion
Solution Approach 1:
The patent transforms the kernel parameters from correlated to orthogonal by applying a transformation matrix. This parameter change maintains implementation feasibility while fundamentally improving convergence stability by eliminating the correlation-induced instability that plagues traditional methods in non-linear environments
Data Source
Figure 1
Figure 2
Figure 3A~3B
AI summary
Disclosed are a method, computing device, and a non-transitory computer-readable medium storing computer-executable instructions which, when executed, cause a processor to perform operations, which may include: sending a transmit signal including non-linear elements; receiving another signal comprising at least a portion of the transmit signal through a receive chain of a computing device; simulating the transmit chain; generating, by the simulated transmit chain, a first signal associated with the transmit signal; generating a second signal by frequency shifting the first signal; generating one or more first kernels associated with the second signal; performing decimation of the one or more first kernels; transforming the one or more first kernels using a lower triangular matrix, into one or more second kernels; combining the one or more second kernels to generate an echo cancellation signal; and subtracting the echo cancellation signal from the received at least a portion of the transmit signal.