Asynchronous Classification of High-Order QAMs

Conventional modulation classification (MC) schemes usually assume that perfect synchronization has been accomplished and then rely on certain statistical characteristics to distinguish different modulation formats. In practice, however, this perfect synchronization assumption is not reasonable since modulation classifiers operate in a non-cooperative manner and therefore, the receiver has little prior knowledge about the transmitted signals and no training is available. In this paper, we first address asynchronous MC for high-order QAMs through blind time synchronization. A characteristic function (CF) based approach is proposed to improve the performance of the conventional cumulant method. We then move on to consider asynchronous MC with frequency offset. We propose a hybrid MC scheme based on blind time synchronization, differential processing, and cumulants to solve this difficult problem.