Multiple random signals are used as excitation sources for identification of multi-variable systems. This paper proposes a new set of the weighting functions of filters which can generate N uncorrelated normal random signals from a single binary random signal.