We consider a blind adaptive multiuser detector which uses minimum output energy (MOE) as a cost function. The MOE detector does not need any information about interference users, and can be implemented blindly using canonical representation. In this paper, we propose an algorithm which has better signal to interference ratio(SIR). The suggested algorithm differs from the MOE algorithm with decision-feedback in a sense that what it uses to estimate the transmitted bit is not filter coefficients of the previous step but convergence characteristics of the MOE algorithm. It has better performance than the MOE algorithm with decision-feedback. The suggested algorithm to estimate the transmitted bit itself can be a detection scheme which does not need the signum function. The simulation verifies that the suggested algorithm has improved SIR.