Superposed multicarrier transmission is a known method to improve frequency utilization efficiency when several wireless systems share the same spectrum. Obviously, an enhanced spectral efficiency comes at the expense of interference. To suppress the effect of interference, forward error correction (FEC) metric masking can be applied. In FEC, the corresponding log-likelihood (LLR) of the superposed band is set to zero or to other proper values determined by the other parameters such as the desired to undesired power ratio (DUR). To be able to apply the FEC metric masking, the information on the superposed band sub-carriers is required at the receiver side. Therefore, in this paper, we propose a novel method for detecting the superposed bands of multicarrier transmissions using Q-learning. We present the simulation results that show a higher rate of superposed band detection accuracy in lower DUR over the conventional method, as well as similar accuracy over other DUR.
