This paper considers a network consists of multiple primary users (PUs) and secondary users (SUs). To seek for the pre- and post-coding matrices of the PUs and SUs that improve the system output, the author applies Max-SINR algorithm for the PUs and cognitive interference alignment (IA) for the SUs. Then by counting the number of variables and the number of equations in the IA condition, this paper analyzes the achievable DoF of all users. Simulation results show that the PUs using max-SINR achieve higher rate than using standard IA with negligible SUs rate reduction.