Long Short-Term Memory based Spectrum Sensing Scheme for Cognitive Radio

Abstract

The application of machine learning models to spectrum sensing in cognitive radio is not uncommon in literature, but most of these models fail to consider temporal dependencies in the signal. In this paper, the temporal correlation among the spectrum data is exploited using a Long Short-Term Memory (LSTM) network. More specifically, the previous sensing event is fed along with the present sensing event to the LSTM model. The proposed sensing scheme is validated based on empirical data of various radio technologies. The proposed LSTM model is compared with other machine learning algorithms in terms of classification accuracy. Furthermore, the proposed scheme is also compared with other spectrum sensing techniques. Results indicate that the proposed scheme improves the detection performance and classification accuracy at low signal-to-noise ratio regimes. Moreover, it is observed that the achieved improvement is obtained at the expense of longer training time and nominal increase in execution time.