Application of Deep Learning Techniques in Blind Identification of Radio Access Technologies

In the last decade, various machine learning schemes have been investigated to make the cognitive radio(CR)more adaptive.Blind identification of radio accesses technology(RAT) indirectly aid the CR to adapt according to the real time wireless environment. In this project work, some of the various wireless standards like GSM, LTE-DL and IEEE 802.11a WLAN are blindly identified using deep neural networks. This report proposes the combination of time-frequency distributions and Convolutional Neural Network (CNN) based Machine Learning technique to identify the RATs. Time-Frequency Analysis (TFA) is used to obtain the spectral content of the signal and AlexNet, a pretrained Convolutional Neural Network is used for feature extraction and identification purpose. The accuracy of the network is analyzed with performance plots of correct identification. Also, performance of the deep neural network classifier has been compared with the previously proposed Machine Learning techniques