This paper proposes a new blind approach to identification of Hammerstein systems, where a static nonlinearity precedes a linear dynamic system. By exploiting input's piece-wise constant property, the denominator of the linear system is identified first together with the order and time delay. Then a modified subspace direct equalization method estimates the unmeasurable intermediate signal, based on which the nonlinearity and linear system are obtained separately. Effectiveness of the proposed approach is demonstrated through a real-time experiment for modeling of magneto-rheological (MR) dampers.