Cognitive Radio (CR) has been proposed as a system-level means to improve the spectrum utilization efficiency. The most significant premise for CR networks is to avoid harmful interference to the licensed users. Therefore, it is very important to estimate the potential interference introduced by the deployment of CR networks so that CR networks can be coexistence with Licensed User (LU) networks by designing proper system parameters. In this paper, two statistical models of potential interference due to CR networks are developed based on transmitter and receiver oriented spectrum sensing schemes, respectively. The approximate probability distributions of aggregate interference introduced by CR networks are then derived with respect to the obtained models. With the help of these probability distributions, a method is further presented to get the critical system parameter, i.e., sensing distance and user density of the CR networks, so that the introduced interference can be controlled to predefined level. Numerical results validate the proposed interference estimation method and confirm the effectiveness of interference distribution based system parameter design.