Approximated MMSE Signal Combining for Interference Rejection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing receive antenna diversity techniques, such as Maximum Ratio Combining (MRC) and Minimum Mean Square Error (MMSE), face challenges in complexity and convergence speed, with MRC being ineffective in inter-cell interference rejection and MMSE requiring complex matrix inversion, while algorithms like Least Mean Square (LMS) and Sample Matrix Inversion (SMI) have limitations in computational intensity and adaptability.
Innovation Solution
An approximated MMSE technique is introduced, which simplifies the calculation of combining coefficients by approximating the correlation matrix as block diagonal, allowing for simpler inversion and initialization of LMS or NLMS procedures, reducing computational complexity and improving convergence speed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If Minimum Mean Square Error (MMSE) combining technique is used, then inter-cell interference rejection performance is improved, but computational complexity increases due to matrix inversion requirements
Solution Approach 1:
The patent segments the full correlation matrix into multiple block diagonal sub-matrices, each corresponding to different antenna groups. This segmentation allows independent inversion of smaller matrices rather than inverting one large matrix, reducing computational complexity while maintaining interference rejection performance.
Solution Approach 2:
The patent changes the parameter representation by using block diagonal approximation of the correlation matrix. This parameter transformation enables the use of simplified inversion techniques on smaller sub-matrices, achieving the same MMSE performance with reduced computational burden.
2Device complexity
If Least Mean Square (LMS) algorithm is used, then computational complexity is reduced, but convergence speed deteriorates
Solution Approach 1:
The patent performs preliminary action by pre-calculating the block diagonal correlation matrix and its inverse before the LMS adaptation process. This preliminary computation provides a better initial weight vector that is closer to the optimal solution, enabling the LMS algorithm to converge faster with fewer iterations.
3Speed
If Sample Matrix Inversion (SMI) algorithm is used, then convergence speed is improved, but computational intensity increases
Solution Approach 1:
The patent applies segmentation by dividing the correlation matrix into block diagonal sub-matrices, allowing parallel or sequential inversion of smaller matrices. This reduces the computational intensity of matrix inversion while maintaining the fast convergence characteristics of SMI-based approaches.
Data Source
AI summary
Signals received from channels exposed to multi-path propagation via a plurality of diversity antennas and including at least one pilot signal are processed by detecting a set of multi-path components for each received signal and computing a set of channel coefficients from multi-path components of the at least one pilot signal in the set of multi-path components. The set of channel coefficients is organized as a channel coefficient vector. From the channel coefficient vector, a set of combining weights is estimated to be applied to the received signals by: computing a spatial correlation matrix of the channel coefficient vector by neglecting the correlations between multi-path components of the channel coefficient vector having different delays, whereby the correlation matrix is a block diagonal matrix including null coefficients other than for non-null sub-matrixes arranged along the diagonal of the correlation matrix, wherein the sub-matrixes have a size equal to the number of diversity antennas; deriving from the spatial correlation matrix a resulting matrix by calculating the inverse of the sub-matrixes or a scaled version thereof; and multiplying the resulting matrix and the channel coefficient vector in order to obtain the desired set of combining weights.