Signal Detection in GSM-Based In-Band Full-Duplex Communication Using DNN

An in-band full-duplex (IBFD) multiple-input multiple-output (MIMO) radio's self-interference (SI) cancellation strength usually determine its performance gains over conventional half-duplex ones. Accordingly, this letter explores an alternative to traditional optimization driven design (ODD) techniques available in literature for beamformer design in IBFD radios. In particular, we consider a generalized spatial modulation (GSM) based bi-directional IBFD system, that provides the flexibility of choosing active antennas among a set of antennas for the spatial symbol. We propose a run-time data-driven prediction approach to solve the multiple multivariate regression problem of detection of the transmitted signal in this GSM based IBFD system in the presence of residual SI arising from non-ideal SI cancellation. We compare the performance of the proposed detector with respect to conventional detectors under several communication parameters that are of practical interest. The proposed DNN-based detector learns the deviations from a standard model without SI and achieves performance that is superior to zero forcing and minimum mean squared error detectors and very close to maximum likelihood detector at faster computation time.