A Novel Transmitter Recognition Model with Fish-SVC Clustering Method

Transmitter recognition with clustering method is an efficient spectrum sensing method to classify different devices and protect the spectrum space security. However, it is hard for the conventional clustering methods to recognize multiple transmitters in non-cooperative communication because some of the methods only process specific spectrum with narrowband. Besides, parameter setting and tuning strategies of the algorithms can be pretty challenging. We propose the recognition model with Fish-SVC clustering method to solve the problems above. We collect wideband spectrum trace data and extract reasonable statistical and physical features to conduct feature vectors of each signal. Our algorithm searches the optimal Gaussian kernel parameter and form reliable clusters. Therefore, we gather the signals that belong to the same transmitters together and recognize the devices. The experiment results show that Fish-SVC can recognize transmitters over wideband spectrum in a non-cooperative communication environment with a nearly 97% accuracy rate.