From the basic conception and the recognition frame of Dempster-Shafer (D-S) evidential theory, a new construction algorithm to basic probability assignment function based on sensor characteristics is proposed. The fuzzy membership for different target identity of the sensing signals from the sensors is processed to obtain the judging precision, which is used as the basic probability assignment function value to perform D-S evidence fusion. An example for the simulations of target recognition focused on three types of sensors is presented, the results show that the method is good to improve the reliability of the target recognition and decreases the uncertainty.