LoRa Signals Classification Through a CS-Based Method

In this paper, a classification method for the identification of the characteristic parameters of an unknown Longe Range (LoRa) signal is proposed. In order to reduce the effective sampling rate, the Compressive Sampling, a new acquisition paradigm that promises of exceeding the Nyquist-Shannon theorem, is used. In particular, values of sampling rate lower than 1Msamples/s have been experienced thanks to a proper random sampling strategy and the exploitation of discrete cosine transform (DCT) to achieve a sparse representation of a LoRa signal. Method performance are assessed by means of MATLAB simulations, using LoRa signals acquired with a proper experimental setup. Normal distributed noise vectors were added to each signal in MATLAB for a broad range of signal-to-noise ratio (SNR) values. As result, the obtained percentage of correct classification for each SNR value assures the reliability of the proposed approach in most operating conditions.