A new algorithm based on joint diagonalization by the householder transformation for convolutive blind separation

Separating convolutive mixtures of source signals is a challenging blind source separation (BSS) problem. A new approach for convolutive mixtures BSS by exploiting the householder transformation to realize the joint diagonalization of high-order cumulant and operating in the frequency domain is proposed. The frequency domain algorithms have a simpler implementation than the time domain algorithms. Joint diagonalization of the fourth-order cumulant in high-order cumulant is operated through using the householder transformation, which can simply and faster implement the joint diagonalization of the matrices. The numerical computer simulations are presented to illustrate the validity and feasibility of our algorithm.