Radar Covariance Estimation for High-Resolution Angle-Doppler Detection

The patent applies parameter changes by modifying the shrinkage estimation parameters and using optimized covariance matrix estimation techniques. Specifically, it employs improved shrinkage intensity parameters and regularization techniques that enhance the accuracy of interference covariance estimation without requiring proportional increases in training data, thereby resolving the contradiction between processing simplicity and estimation efficiency

The patent uses copying by creating refined copies of the covariance matrix through multiple estimation passes and ensemble methods. It generates improved covariance estimates by combining multiple smaller training sets through techniques like sample averaging and bootstrap resampling, achieving reliable interference characterization with fewer total training samples while maintaining processing efficiency

The patent applies segmentation by dividing the training data into multiple smaller, independent subsets that are processed separately through parallel covariance estimation. This segmentation allows the system to achieve accurate interference covariance estimation by combining results from multiple small training sets, eliminating the requirement for large monolithic training datasets and thereby improving system efficiency while maintaining estimation accuracy

The patent employs dimensionality change by transforming the estimation problem from requiring many samples in the time domain to achieving equivalent accuracy through ensemble averaging across multiple independent realizations. This approach changes the dimensional requirements from sample size to ensemble size, allowing efficient processing while maintaining measurement precision