Performance Analysis of LSTM-CNN for Spectrum Sensing in Cognitive Radio Networks

Abstract

Spectrum sensingis a primary task for Cognitive Radio Networks. Deep Learning Models have proven its efficiency in SS and currently lot of research is focused on implementing it in practical Scenarios. However, for practical implementation, it is necessary that spectrum sensing should be from assumptions that are made by deep learning models. CNN is a very powerful model for extracting spatial characteristics such as sample covariance matrix. On the other hand, LSTM uses predictive sensing by extracting time series data. For efficient Spectrum sensing -Deep learning model, this paper proposes LSTM-CNN model to extract temporal as well as spatial data from the incoming signal. According to simulation results, LSTM-CNN outperforms CNN and the LSTM method separately.To further prove efficiency of the model we have compared the result from ML techniques like SVM and LR also.