Data Collection for Classification of Radio Signals Using SDR Transceivers

The paper deals with the problem of automatic classification of telecommunication signal modulation types using machine-learning methods. The task is actual in the field radio electronic, with the advent of integrated telecommunication systems, new modulation, and coding methods, it became necessary to search and develop simple, efficient, reliable methods for processing signals in channels with interference. The type of modulation is an important parameter on which the signal processing circuit of the receiving or transmitting device depends. It is impossible to extract useful information from the received signal without knowing the type of modulation. A database for the problem solution was collected through the experiments with software-defined radio. During the experiments the signals at a frequency of 900 MHz, with modulations: BPSK, GMSK, NBFM, OFDM were collected. The most popular of the contemporary methods for classification are evaluated and compared. As the result of this study, a method for collecting data using an SDR transceiver was considered, and a comparison was made of the accuracy of algorithms trained on the collected dataset.