Multisensor Modulation Classification (MMC): Implementation Considerations -- USRP Case Study

This paper provides hardware implementation considerations for previously developed algorithms designed to improve the classification of the modulation of weak radio signals utilizing multiple sensors. The case study presented focuses on a likelihood-based approach in a centralized data fusion framework. Data sets from multiple sensors are fused to obtain a more accurate modulation classification as previously demonstrated in simulations. The algorithms are implemented on a hardware test bed that consists of the laboratory grade software defined radio platforms. The performance is examined in realistic environments and compared with results obtained via simulations. The test bed results indicate that the predicted performance improvements are difficult to achieve in practice and the algorithms need to be tailored to account for hardware features and signal propagation effects. Differences between results obtained in simulations and in hardware implementation are discussed and adjustments are made to achieve consistent improvement necessary for refinement of the solution toward military applications.