Dictionary reduction: A preprocessing method for efficient sparse reconstruction-based direction-of-arrival estimation

Direction-of-arrival (DOA) estimation based on sparse reconstruction typically constructs an overcomplete dictionary over the array observable range. Due to source sparsity, most dictionary atoms are redundant. This paper proposes two simple dictionary reduction methods to reduce computational cost. Method 1 calculates the distances of each atom to the signal subspace and the noise subspace and retains those atoms that are closer to the signal subspace. Method 2 reduces the dimension of either the signal subspace or the noise subspace to make the two subspaces equal in dimension and then performs dictionary reduction based on these two subspaces. Numerical simulations show that combining dictionary reduction with classical sparse reconstruction methods significantly reduces computational cost while also enhancing DOA estimation accuracy and resolution. Experimental results further validate the effectiveness of the proposed methods.