This paper introduces a novel method to assess the single step balance and detect unexpected falls based on whether the swing foot could chase up the capture point in a single step. The method grounds on the fact that the Linear Inverted Pendulum Model (LIPM) with finite size foot can describe the dynamic walking procedure of human accurately, which is validated in our study. Then the relationship between the position of swing foot and capture point is analyzed, based on which the single step balance assessment method is presented. To verify the performance of proposed method, a serials of falling experiments is conducted. Results confirmed that the implemented method allows us to well estimate the balance situation and rapidly detect potential falls with a low false alarm rate when human walks in low and medium speed. Moreover, our approach works well in walking and standing, as well as other transitory periods without any falling-based training, so it seems promising to be applied to assistance devices to help people regain balance when facing unexpected perturbations.