Modulation Classification in Multipath Environments Using Deterministic Particle Filtering

We address the challenge of modulation classification in an unknown dispersive environment. A bank of deterministic particle filters (DPF) is used to jointly estimate the communication channel and data sequence for each possible modulation scheme, and the best path metrics from each DPF form a feature vector. Maximum likelihood modulation classification is performed and makes use of the statistics of the feature vector under different modulation schemes. Simulation results show that the algorithm can successfully classify the observed modulation schemes with as few as 50 symbol observations, making the algorithm practical under time-varying conditions, as well. Additionally, since the communication channel and the transmitted sequence are estimated in the modulation classification process, the proposed algorithm is a natural choice for joint channel estimation, data detection, and modulation classification.