Extended state observer (ESO), which works with the augmented state variable of the nonlinear system, is proposed and employed to straightforwardly reject disturbances by the online disturbance reconstruction (estimation) and attenuation technique under the active disturbance rejection control (ADRC) scheme. Focused on more precise disturbance estimation on unknown high orderperturbation signals or fast-varying disturbances, which are difficult to be dealt with by most of traditional ESO, yet widely emerging in practical application control systems, a novel global robust finite-time ESO is designed and expounded for rapidly and accurately tracking such internal and external disturbances. The proposed approach employs finite-time control theory and compensation control method by means of the sliding mode switching term to ensure that the state variable estimation error can converge in finite time to zero for various forms of system disturbances. The corresponding stability analysis is presented in terms of the Lyapunov method. In addition, the validity of the robust finite-timeESO is verified via simulations and revealed to be more effective than the traditional linear and nonlinear extended state observer.