The extended state observer (ESO) is used to estimate both the unmeasurable system states and acting lumped disturbance. Generally, high gains need to be selected in ESO to achieve fast convergence, which can make it sensitive to measurement noise. To address this limitation, a governing structure with a special combination of Kalman filter (KF) and ESO is proposed. The former serves as noise filtration while the latter is responsible for on-line reconstruction of states and disturbance. The two are connected as the system model used for the KF design contains a constantly updated estimate of the lumped disturbance obtained with the ESO. Finally, the resultant estimates are used to construct a composite disturbance rejection-based controller. To verify the new method, hardware experiments are performed on an electronic throttle system. Conducted quantitative comparison with conventional solution reveals advantages of the proposed approach in noise attenuation and control performance.