Efficient independent component analysis with reference algorithm

Regarding to the relative slow convergence of traditional independent component analysis with reference (ICA-R) methods, an improved ICA-R algorithm is proposed referring to a novel objective function, which is derived by adding the reciprocal of similarity measure to the standard contrast function, then the Lagrange multiplier method is adopted on the novel objective function, and the optimal weighted vector is obtained efficiently. As a result, the interested source signals can be extracted by a special linear transformation. The proposed improved ICA-R algorithm not only can avoid ineffective inequality constraint, but also has a faster convergence speed and higher extracted quality compared with state-of-the-art ICA-R methods. Simulation results show that the proposed algorithm is able to extract the desired source signals and yield good performance.