This paper estimated probabilities of bankruptcy for U.S.  firms from 1996 to 2007 in a model where common equity is viewed as a down-and-out call option(DOC) on a firm’s asset (Brockman and Turtle, 2003). It recognizes that a bankruptcy or default may happen anytime rather than only on the days when debt matures. In structural credit risk models, we rely on equity price information to back out the market value and volatility of firm’s asset. Yet observed equity prices can diverge from their equilibrium values due to microstructure noise. Ignoring trading noise could lead to bias in the estimates of default probabilities. To address the issue of possible trading noise, we adopt a smoothed particle filter technique in conducting MLE estimation (Duan and Fulop(2007) ) of our DOC model.  Physical (real) probability of defaults for corporate are constructed. CRSP and Compustat Database data are used. Monte Carlo simulation is implemented to check the reliability of this approach. This new approach was also compared with other commonly adopted bankruptcy models, such as Altman (1968), Merton(1974)/ KMV  and Shumway(2001).