Joint diagonalization learning algorithm for nonlinear blind source separation

Summary form only given. Recovering independent source signals from their nonlinear mixtures is a very important issue in many fields. This paper proposes a robust radial basis function network (RBFN) approach by using higher-order cumulants (HOC) when observations suffer from noise and nonlinear distortion. The HOC can measure the departure of a random vector from a Gaussian random vector for extracting the non-Gaussian part of a signal. The proposed method can efficiently recover nonlinearly mixed signals suffered high-level noise simultaneously. It is divided into two steps. First, the RBFN helps us to transform the mixed signals to high-D space. Then in the second step, we can linearly separate the mixtures in the high-D space by joint approximate diagonalization of eigenmatrices. We consider artificial signal and acoustic signal separation and denoising applications. Furthermore, a comparison between the traditional RBF-based method, original JADE, and proposed algorithm is produced, from which we can see the proposed algorithm is more suitable and applicable for unsupervised nonlinear signal denoising problems.