基于深度学习的频谱感知研究综述

Survey on Deep Learning based Spectrum Sensing

频谱感知可以提高认知无线电网络的频谱利用率,但传统的频谱感知方法不能在复杂的通信环境中进行快速的频谱感知。因此,借助计算机计算能力的提升,将深度学习应用于频谱感知,以快速、智能地获得感知结果。首先,介绍在频谱感知中应用较为广泛的深度学习模型,包括卷积神经网络、长短期记忆网络和深度强化学习;其次,对近几年基于深度学习频谱感知的研究成果进行综述,包括基于卷积神经网络的频谱感知、基于长短期记忆网络的频谱感知、基于深度强化学习的频谱感知和利用其他深度学习模型的频谱感知方法;最后,对当前深度学习频谱感知方法存在的问题进行思考,展望未来的研究方向。

Spectrum sensing can improve the spectrum utilization of cognitive radio system,but traditional spectrum sensing methods cannot perform spectrum sensing rapidly in the complex environment.With the improvement of computer computing power,deep learning can be applied to spectrum sensing to obtain the sensing results quickly and intelligently.Firstly,deep learning models which are widely used in spectrum sensing are introduced,including convolutional neural network,long shortterm memory network and deep reinforcement learning.Then the research results of spectrum sensing based on deep learning in recent years are summarized,including spectrum sensing based on convolutional neural network,spectrum sensing based on long short-term memory network,spectrum sensing based on deep reinforcement learning and spectrum sensing methods using other deep learning models.Finally,the problems existing in current deep learning spectrum sensing methods are considered and the future research directions are prospected.