Accurate Automatic Extraction of Signal Components from Noisy Radio Spectrograms

In some radio communication scenarios it is useful to extract the bandwidth and start/end times of each transmission in a spectrogram. However, few methods in the literature are able to extract this information without human manual intervention, and the few existing ones have a limited accuracy. In this context, this work overcomes these limitations by proposing a novel fully automated method that can provide this information with high accuracy. The results obtained from simulations and hardware experiments show that the proposed method outperforms other methods available in the literature, achieving a virtually perfect accuracy for signal-to-noise ratio values as low as -10 dB.