An augmented Sparse Iterative Covariance-based Estimation Method based on Elastic Net for DOA Estimation
Zhengxiang Wu, Ling He, Xiao Ling Yan, Qian Wang, Lanfeng Xie · 2019 IEEE International Conference on Signal, Information and Data Processing (ICSIDP) · 2019
In this paper, an innovative SPICE approach based on elastic net model, abbreviated as EN-SPICE, is presented, for array direction of arrival (DOA) estimation. It is shown that the elastic net model can be exploited to establish a new sparse representation with two kinds of weighted norm, which is not just restricted to the mixture of ℓ1-penalty and ℓ2-penalty. This method can bring to a grouping effect in the process of variable selection. Distinguishing from the previous sparse covariance-based estimation method, the EN-SPICE algorithm adopts a scaling transformation to preserve the variable selection property and is one of the simplest ways to restrain shrinkage, especially under the circumstance of the number of predictors is much bigger than the number of observations. Consequently, the SPICE based on elastic net can increase the sparsity level of variable, namely improving the precision of variables. Through the covariance matrix of signal computation to reconstruct the signal features, one can perform the signal parameter estimation. In the latter part of this paper, Monte Carlo simulation results demonstrate that the proposed method can achieve much better estimation accuracy than other SPICE algorithm and the iterative loop is much more effective.