Automatic Identification of MIMO-STBC Signals via Fusion of Multiple Characteristics

Signal identification, a fundamental technology in cognitive radio networks, plays a crucial role in selecting the optimal spectrum to ensure efficient spectrum utilization. Most current algorithms overlook the fact that the noise distribution model varies. This leads to significant identification performance degradation. In this paper, a novel signal identification scheme is proposed for MIMO-STBC systems to introduce an unknown noise elimination mechanism and goodness-of-fit feature fusion to address the degradation caused by varying noise models. First, the generalized linear norm is leveraged to construct the generalized linear regularized denoising optimization problem. Subsequently, the goodness-of-fit features of the signal are analyzed and integrated in the presence of residual noise. Finally, a novel maximum likelihood fusion classifier is developed, which introduces a weighted averaging criterion, and multiple weighted factors are integrated with the log-likelihood function to fuse multi-source and multi-feature information. Simulation results indicate that the proposed algorithm exhibits excellent recognition performance in the presence of unknown noise, achieving over 95% accuracy for the four MIMO-STBC signals at an GSNR of 6dB. Additionally, the algorithm requires no priori information, such as channel coefficients, the number of transmit antennas, and signal power.