Cooperative sensing is one solution for fading, hidden node problems, and uncertainty noises in primary signal detection. In this case, each secondary user (SU) senses available channel and exchanges their local sensing results to decide which channel can be accessed. Subsequently, this technique is capable to reduce sensing errors and lead to the throughput improvement. In this paper, we investigate decentralized cooperative user (DCU) with limited sensing capability under Partially Observable Markov Decision Process (POMDP). It models license spectrum opportunities as Markov discrete process where the idle and busy state of spectrum is represented by 0 and 1. Each SU with limited sensing capability detect primary channel under the constraints of collision allowable by primary user (PU) and reduce the interference to PU activity. SUs exchanges their sensing results based on the local observation. To validate this work, the performance of DCU is compared with single user sensing and random cooperative user (RCU). © 2012 IEEE.