A blind receiver for PCMA signals aims to detect the data symbols of the two carrier signals contained in the received aggregate signal. In order to execute a multi-user detection or sequence estimation scheme, the receiver needs to obtain knowledge on several signal and channel parameters. With the symbol rate, carrier frequencies and modulation schemes assumed to be known from prior processing steps, a blind estimation method for the powers of the carrier signals, and their respective individual signal-to-noise ratios, is proposed in this paper. The approach is based on the estimation of statistical features of the received signal that are selective w.r.t. the noise and the respective carrier signals by exploiting a sufficient deviation of their frequency offsets from each other. In particular, the cyclic correlation and the cyclic cumulants are estimated and used to infer the powers. Numerical simulation results demonstrate the usefulness of the proposed scheme.