Algorithm for underdetermined blind signal separation based on constrained NMF
基于约束NMF的欠定盲信号分离算法*

A constrained nonnegative matrix factorization (NMF) method is proposed to resolve the problem of underdetermined blind signal separation in this paper. It is hard to obtain unique factorization and correctly separate source signals when NMF is directly applied to resolve problem above. In this paper, on the basis of standard NMF, determinant criterion was imposed on estimated mixing matrix to achieve the unique factorization of NMF; sparsity and the least correlated component constraints was imposed on estimated sources to realize the unique separation of mixed signals and improve the performance of source separation. Simulation results show effectiveness of the proposed algorithm.