Automatic modulation classification of radio signals by using sensor networks and machine learning techniques

By using a network of cooperative sensors, distributed signal reception is enabled and thus a significant improvement of automatic modulation classification (AMC) performance is achieved, specially in case of radio signal reception via the multi-path channels. AMC models based on five machine learning techniques (Naive Bayes, Support Vector Machines, ADA Boost, Random Forest and Artificial Neural Network) were developed, while the classification was performed based on the estimation of the fourth-order cumulant with the signal reception over mutually independent multi-path channels. A scenario with the random distribution of measuring receivers in space around the transmitter is observed. The results of the analysis show a significant gains in terms of classification accuracy.