We present an optimum diversity receiver called blind ratio combining (BRC) that minimizes the average symbol error probability or maximizes the average output SNR, where the channels' time delays and the random phases are known, while the fading amplitudes are unknown. In contrast to previous works, where efforts were made to find a posteriori probabilities at the receiver, the BRC simply calculates the optimum weights, which depend on the channel's statistics, avoiding continuous channel estimation, and thus, it significantly reduces the system's complexity. In nonidentical multipath fading channels with power delay profile (PDP), the BRC receiver performs between maximal ratio combining (MRC) and equal gain combining (EGC), and keeps its performance comparable - and in some cases superior - to that of generalized selection combining, while for large values of the decay factor, it approaches MRC. Moreover, in the important practical case of exponential PDP - common in RAKE receivers modeling and adopted for the Universal Mobile Telecommunications System spatial channel modeling by the European Telecommunications Standards Institute-3GPP - the optimum weights can be accurately approximated by simple elementary functions. Furthermore, it is proved that the utilization of these weights ensures an error performance improvement over EGC for arbitrary PDPs. The proposed BRC receiver can be efficiently applied in wireless wideband communication systems, where a large number of diversity branches exists, due to the strong multipath effects.