The ability to answer all important questions about the radio-frequency (RF) scene is essential for cognitive radios (CRs) to be effective. In this paper, we propose a RF -based automatic traffic recognizer that, observing the radio spectrum emitted by a communication link and exploiting machine learning (ML) techniques, is able to distinguish between two types of data streams. Numerical results based on real waveforms collected by a RF sensor, demonstrate that over-the-air user traffic classification is possible with an accuracy of 97% at high signal-to-noise ratios (SNRs). Moreover, we show that using a neural network (NN) very good classification performance can be achieved also at low SNRs (around 2 dB). Finally, the impact of the observed RF bandwidth and the acquisition time window on the classification accuracy are analyzed in detail.

Machine Learning for User Traffic Classification in Wireless Systems

TESTI, ENRICO;FAVARELLI, ELIA;Andrea Giorgetti
2018

Abstract

The ability to answer all important questions about the radio-frequency (RF) scene is essential for cognitive radios (CRs) to be effective. In this paper, we propose a RF -based automatic traffic recognizer that, observing the radio spectrum emitted by a communication link and exploiting machine learning (ML) techniques, is able to distinguish between two types of data streams. Numerical results based on real waveforms collected by a RF sensor, demonstrate that over-the-air user traffic classification is possible with an accuracy of 97% at high signal-to-noise ratios (SNRs). Moreover, we show that using a neural network (NN) very good classification performance can be achieved also at low SNRs (around 2 dB). Finally, the impact of the observed RF bandwidth and the acquisition time window on the classification accuracy are analyzed in detail.
2018 26th European Signal Processing Conference (EUSIPCO)
2040
2044
Enrico Testi ; Elia Favarelli ; Andrea Giorgetti
File in questo prodotto:
File Dimensione Formato  
Eusipco2018_postprint.pdf

accesso aperto

Tipo: Postprint
Licenza: Licenza per accesso libero gratuito
Dimensione 498.1 kB
Formato Adobe PDF
498.1 kB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11585/674342
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 15
  • ???jsp.display-item.citation.isi??? 12
social impact