Cooperative spectrum sensing (CSS) was proposed as one solution to overcome uncertainty noises in primary signal detection. It increases capability of primary signal detection and accuracy. This paper investigates the performance of decentralized CSS model with limited sensing capability in cognitive radio (CR) system. Simulation considers multiple secondary users (SUs) sense license channel and distributes their sensing result among others. Partially observable Markov decision process (POMDP) framework is used as an aid for SU that partially sense over license channel. It models license channel as Markov discrete process where idle and busy states is represented by number of �0� and �1�. To validate this work, we make some comparisons against single user sensing, centralized multiuser cooperative spectrum sensing (CMCSS) and decentralized random CSS model. The derived results show that CMCSS and DMCSS model has better performance than single user sensing and decentralized random CSS model. CMCSS outperforms DMCSS model. However, DMCSS model still outperforms decentralized random CSS. At certain case, DMCSS performance matched to the throughput achieved by CMCSS when probability of selected channel (p) = 1. © 2015 School of Electrical Engineering and Informatics. All rights reserved.