A new state estimation scheme is presented for multidimensional dynamic systems. The state vector is estimated, component-by-component, by the parallel use of the stack sequential decoding algorithms of information theory. A faster state estimation is accomplished with this scheme for longer time intervals with available memory. Simulation results, some of which are presented, have shown that the scheme produces better estimates than the Kalman estimates for some non-linear models with white gaussian observation and disturbance noise, while it produces estimates comparable with the Kalman estimates for linear models with white gaussian noise.