Abstract:For the case of larger condition number of mixing matrix in the underdetermined linear instantaneous mixtures, the noise amplification problem caused by the source signals was solved with the method of Least Squares (LS) Estimation, and an algorithm for solving the separation matrix was investigated. For this method, it establishes cost function based on the principle of maximizing signal to interference plus noise power ratio (SINR), and then the separation matrix is derived by optimizing this cost function. Simulation results show that, compared to the LS method, the proposed algorithm can improve the separation performance under noise amplification with larger condition number of mixing matrix.