A Bayesian Approach for Adaptively Modulated Signals Recognition in Next-Generation Communications

By promoting spectrum efficiency and transmission reliability, link adaptation (e.g., adaptive modulation) is one of the enabling technologies for next-generation 5G communications. In this paper, we investigate the recognition of adaptively modulated signals, which remains still as an unexploited area as far as we are aware, especially in the presence of time-varying fading channels. A unified model, relying on the dynamic state-space approach, is formulated, which thoroughly characterizes the coupling relationship between two hidden states, i.e., unknown modulation schemes and fading channels. In contrast to existing schemes marginalizing directly out random fading effects, a joint estimation paradigm, which relies on the Bayesian stochastic inference and a maximum a posteriori criterion, is developed to acquire time-correlated fading states sequentially, at the same time of recognizing unknown modulation schemes. In order to alleviate the computation complexity, two simplified schemes, i.e., fading-driven and goal-oriented, are designed. It is demonstrated that, by fully exploiting the underlying dynamics of the estimated fading gain which is modeled by a discrete-states Markov chain, the recognition performance of adaptive modulations will be improved significantly. The proposed system model and a sequential estimation framework, by providing additionally the dynamic fading channels, may be of great promise to more flexible and effective link adaptations.