Gaussian Pseudorandom Number Generator Based on Cyclic Rotations of Linear Feedback Shift Registers

Abstract

This paper presents a new proposal to generate optimal pseudorandom numbers with Gaussian distribution. The generator is especially designed for low-cost hardware implementation, although the software version is also considered. For this reason, Linear Feedback Shift Registers in conjunction with cyclic rotations are employed. The proposal presents a low implementation cost and overcomes the limitations of the previous Gaussian generators based on linear feedback shift registers by means of a less complex algorithm to find optimal configurations. As a consequence, it turns into a really usable generator. Moreover, a further improvement, based on the simulated annealing algorithm, is applied in order for the random values to be better adjusted to the normal distribution.