基于注意力机制的深度学习频谱感知方法

Spectrum Sensing Based on Convolutional Neural Network and Attention Mechanism

深度学习具有出色的自动特征学习能力,比传统的机器学习方法具有更好的性能。注意力机制可以给予局部焦点更多的关注,而且还可以通过过滤掉无用的信息来降低计算复杂度。因此,具有注意力机制的深度学习可以有效实现自动特征学习,以及降低计算复杂度。本文针对认知无线电系统中主用户信号随机到达与离开时的频谱感知问题,提出了一种结合注意力机制的深度学习的感知方法。仿真结果表明,相比其它感知方法,所提出的频谱感知方法能够在主用户信号随机到达与离开的情况下有效工作及表现出优越的性能。

Deep learning has excellent automatic feature learning capabilities and has better performance than traditional machine learning methods. The attention mechanism can give more attention to the local focus, and it can also reduce the computational complexity by filtering out useless information. Therefore, deep learning with attention mechanism can effectively realize automatic feature learning and reduce computational complexity. Aiming at the problem of spectrum sensing when the primary user signal arrives and departs randomly in the cognitive radio system, this paper proposes a deep learning sensing method combined with the attention mechanism. The simulation results show that, compared with other sensing methods, the proposed spectrum sensing method can work effectively and exhibit superior performance when the primary user signal arrives and departs randomly.