CNN-based Burst Signal Detection with Covariance Matrix

One of the critical issues in burst signal detection is to design the optimal sensing duration. Conventional energy detection methods generally determine the threshold based on the target detection performance, such as false-alarm probability or detection probability. However, these approaches could not accurately detect the burst signal without prior information. Motivated by this, in this paper, we use a convolutional neural network (CNN) to effectively extract the feature from the data. As a realization of the developed CNN-based detector, we adopt the sample covariance matrix as the input data of a neural network. Finally, we also show the performance of the proposed method for both optimal and non-optimal sensing cases. Particularly, the proposed method could achieve a detection probability of 98.4 % for the non-optimal sensing case at SNR= −15 dB, which significantly outperforms the conventional methods.