Matrix Recovery Algorithm Based on Normal Vector of Hyperplane in Underdetermined BSS

摘要

This paper discusses the matrix recovery for sparse source under the k-SCA condition. Here, to estimate the mixing matrix using hyperplane clustering, we propose a new algorithm based on normal vector for hyperplane. The algorithm is an extension of blind separation algorithm using statistical clustering (i.e. kmeans). Compared with the Hough SCA algorithm, we give a method to calculate normal vector for hyperplane, and our algorithm has lower complexity and higher precision. Two examples demonstrates its performance.