Receiver Filtering for Large-Antenna Noise Covariance Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Large antenna systems face challenges in accurately estimating noise covariance matrices, leading to ill-conditioned systems and faulty data decoding due to under-dimensioning, particularly with 64 receive antennas where the 64*64 matrix is under-determined.
Innovation Solution
A method and system for receiving signal streams that involve estimating channels for each antenna, grouping received signals, performing group-specific filtering, and applying a second filtering step to produce filtered signals, which are then processed to improve data decoding accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If noise covariance matrix estimation is performed for large antenna systems (e.g., 64 receive antennas), then the system attempts to handle large-scale MIMO communication, but the estimation becomes under-determined and ill-conditioned due to insufficient training elements
Solution Approach 1:
The patent divides the 64 receive antennas into multiple sub-groups (e.g., 4 sub-groups of 16 antennas each). For each sub-group, a separate noise covariance matrix is estimated using only the training elements allocated to that sub-group. This segmentation transforms the under-determined 64x64 matrix estimation problem into multiple well-determined smaller matrix estimation problems, where each sub-group has sufficient training elements relative to its antenna count, thereby improving estimation accuracy and avoiding ill-conditioning.
2Measurement precision
If a single noise covariance matrix is estimated for all antennas using limited training elements, then the system complexity remains low, but the estimation becomes ill-conditioned leading to faulty decoding
Solution Approach 1:
The equalizer design is segmented into multiple independent noise covariance estimation processes, one for each antenna sub-group. Each sub-group's noise covariance matrix is estimated separately using local training elements, and these estimations can be performed in parallel. This approach improves measurement precision by ensuring each estimation has sufficient data, while managing complexity through modular, parallelizable processing units.
Solution Approach 2:
The patent transitions from estimating a single high-dimensional 64x64 noise covariance matrix to estimating multiple lower-dimensional sub-group matrices. This dimensional transformation reduces the complexity of each individual estimation problem while collectively covering all antennas. The problem moves from one high-dimensional space to multiple lower-dimensional spaces, making each estimation well-conditioned and solvable with available training elements.
Data Source
AI summary
Embodiments of the present disclosure are related to a receiver and a method for receiving signal. The method comprises estimating, by a receiver, channel for each of a plurality of antennas using the received signal stream, to produce a plurality of estimated channels. Also, method comprises grouping a predetermined number of received signals from the received signal stream and estimated channels associated with the received signals to obtain a plurality of groups. Further, a group specific filtering is performed on the received signals and corresponding estimated channels of each of the plurality of groups to obtain filtered received signals and a plurality of filtered estimated channels. Thereafter, the method comprises performing second filtering on all the filtered received signals to produce second filtered signals, which are processed to produce an output.