Cooperation among secondary users can greatly improve the spectrum sensing performance in cognitive radio. In our previous work , we proposed the diversity as a measure of the cooperative gain and developed cooperative spectrum sensing schemes based on some pre-defined local threshold selections. In this paper, on the other hand, we start the cooperative spectrum sensing development from the fusion rule selection. The detailed analysis of two relatively simple and mathematically tractable rules, namely the AND and the OR rules, is provided in this paper. Interestingly, it is found that the AND rule results in the same performance as the non-cooperative sensing strategy and thus cannot achieve diversity gain at all. On the other hand, it is shown that when SNR is large, the OR rule is optimum according to the large deviation analysis  and proved to achieve the maximum diversity gain with better performance than our previous sensing scheme developed in . Based on the fixed decision rule, we find the optimal local sensing threshold accordingly. Numerical results are provided to verify our analyses.