Cooperative spectrum sensing can solve the problems such as channel fading or hidden terminal in local spectrum sensing of cognitive radio network effectively. This paper researched the weighted cooperative spectrum sensing technology and proposed a reliability-based cooperative spectrum sensing algorithm. In the algorithm, the local spectrum decision is obtained with the maximum-eigenvalue-based detection scheme in each sensing user. Then it is compared with the global decision and the reliability of each sensing user is estimated in fusion center. The fusion center reckons the credibility threshold using of the Trimmed Mean and selects the users which reliabilities are larger than the credibility threshold to participate in the cooperative spectrum sensing. This algorithm reduces the complexity of the system, improves the performance of the spectrum sensing, and is robust to the noise uncertainty. Simulation results showed that the proposed algorithm outperforms other weighted algorithms 1~3 dB in signal to noise ratio and save the communication overhead 43.75% or so.