In the blind FIR system identification, we have an advantage of the robustness to the Gaussian noise, when using only the 3rd and 4th order cumulants. However, the identification error increases when it has a singularity which is mainly due to the overestimation of the system order. It is the result of the fact that the overestimation of the order doesn't guarantee the independence of FIR system coefficients generated from variable parts of the 2nd order system and those of the 3rd order systems. In this thesis, we solved a least square problem with singular cumulants matrix using the concepts of 6th order Euclidian distance. Using computer simulations, we showed that it suppresses the dependence among system coefficients and decreases identification errors.
