Asynchronous Network Signal Sensing and Classification Techniques

This SBIR project will develop a distributed sensor network framework capable of synchronous/asynchronous automatic modulation classification (AMC) in a non-cooperative communication environment where the received signal is unknown and weak. During the Phase I of this effort, we will focus our attention to synchronous/asynchronous AMC based on a distributed decision fusion network using parallel fusion architecture, because of the robustness, the scalability and the flexibility it offers. The developed distributed sensor network will include optimized local sensor decision rules as well as optimized fusion rules at the fusion center. The first goal of this project will be the generalization of the well-established distributed detection and data fusion results to the multiclass weak signal AMC problem which is not a trivial task. We will investigate both soft decision based multiclass classifiers along with multiclass fusion rules and binary decision tree classifiers. We will also address various other open research problems in distributed AMC including incorporation of the wireless channel uncertainties between local sensors and the fusion center, bandwidth management in a bandwidth constrained sensor network and security issues. We will perform feasibility studies in a number of operating conditions through computer simulations and theoretical analyses.