Cognitive radio is gaining increasingly interest as a promising solution to current spectrum resource shortage. Within various tasks of cognitive radio, spectrum sensing is the fundamental one, but is challenged by wireless channel fading. By collecting diversity among different users, cooperative sensing can overcome the fading problem very well. Usually, only the local binary decisions are available for sensing cooperation due to limitation of the channel bandwidth. However, in our previous work , we have shown that this strategy will either sacrifice diversity or signal-to-noise ratio (SNR) gain. In this paper, we will study cooperative sensing with ternary local decisions. Compared with the binary cooperative sensing, this strategy will can regain diversity and recover the extra SNR loss by appropriate threshold selection, without increasing the decision forwarding bandwidth.