The simultaneous parameter identification and state estimation algorithm in the linear discrete-time system described by the canonical form is developed in computational purpose. The problem is formulated under the assumption that the system parameters are unknown constants and the noise statistics are given a priori as well as the known order of the system. The procedure requires the Kalman filter which is applied for state estimation and parameter identification at each stage, and the smoother which is for state-smoothing. The smoothing calculation is utilized to get the observation matrix with respect to parameter vector. Two approaches where the parameter has fictitious noise inputs and nothing are shown. The effective choice of fictitious noise has led the estimation better in the numerical example.