In this paper, we study the problem of improving capacity of multiuser OFDM - FDMA cognitive radio (CR) system. We propose new algorithms, namedly Interference Inversion Allocation (IIA) and Quantized-Interference Inversion Allocation (Q-IIA) to allocate subcarrier to users of the multiuser CR system. These techniques allocate subcarrier to users inverserly proportional to its interference to primary user (PU). In addition, we also propose to use windowing technique, a technique being used widely in 802.11 family(WiFi) systems, to multiuser CR system to improve its capacity. The computer simulation is carried out for comparison for three cases of allocation such as IIA, Q-IIA and uniform subcarrier allocation, both with and without windowing. The simulation results show that new algorithms IIA, Q-IIA provide better system capacity than uniform subcarrier allocation does. Among IIA and Q-IIA algorithms, IIA brings system capacity higher than Q-IIA does. Relating windowing technique, the multiuser CR system can achieve much higher capacity (2.6 times)in case of using windowing than without using it, applicable to all IIA, Q-IIA, and uniform subcarrier allocation cases.