Abstract: A new robust minimum variance filtering method for uncertain linear discrete-time systems is developed, which can guarantee minimizing bounded variance of filtering error. Unlike other robust filters, this filter only needs to solve two independent Riccati difference equations. So it is simple and effective to calculate and analyze. The existence of the new filter can be ensured by the quadratic stable of the sub-system that is affected by uncertain parameters. This result relaxes the assumption of quadratic stable of the whole system, which is required for some other robust filters. The practical calculation shows that the new robust Kalman filter is more effective.