Data model extraction - a new method of signal detection for multi-level index modulation

Instead of treating the data as no model or a fixed model, can we detect the data through modeling it dynamically for wireless communications? In this paper, a new method, data model extraction, for signal detection is studied for multi-level index modulation (MTL-IM) system. The reason why MTL-IM is investigated is based on its natural attributes that the information symbols are divided into explicit categories by distinct energy. A one-size-fits-all continuous model rather than a discrete model is used to describe the data. In this way, the problem of estimation of the discrete values of the energy level is transformed into the problem of estimation of both the constructed model itself and its parameters. Subsequently, the optimization function is built over the constructed model of the data. By this means, a mechanism of picking up information through model matching is established. Also, a new data-model-extraction-assisted adaptive message delay passing algorithm, in which the learning rate is introduced to delay the message from the adjacent nodes, is investigated. Simulation results demonstrate the effectiveness of the data model extraction method and the proposed solution.