Statistical Signal Transmission (SST) is a technique based on orthogonal fre-quency-division multiplexing (OFDM) and adopts cyclostationary features, which can transmit extra information without additional bandwidth. However, the more complicated environment in 5G communication systems, especially the fast time-varying scenarios, will dramatically degrade the performance of the SST. In this paper, we propose a fragmen-tal weight-conservation combining (FWCC) scheme for SST, to overcome its performance degradation under fast time-varying channels. The proposed FWCC scheme consists of three phases: 1) incise the received OFDM stream into pieces; 2) endue different weights for fine and contaminated pieces, respectively; 3) combine cyclic autocorrelation function ener-gies of all the pieces; and 4) compute the final feature and demodulate data of SST. Through these procedures above, the detection accura-cy of SST will be theoretically refined under fast time-varying channels. Such an inference is confirmed through numerical results in this paper. It is demonstrated that the BER perfor-mance of proposed scheme outperforms that of the original scheme both in ideal channel estimation conditions and in imperfect chan-nel estimation conditions. In addition, we also find the experiential optimal weight distribu-tion strategy for the proposed FWCC scheme, which facilitates practical applications.