基于卷积神经网络的自动调制识别方法

Automatic modulation recognition method based on convolutional neural network

针对调制信号识别精度不足的问题,提出了基于卷积神经网络和注意力机制的识别方法。该方法在卷积神经网络的卷积层与池化层之间增加了注意力机制,赋予调制信号关键特征更高的权重,对信号进行时频分析并转换为频谱图作为模型的输入,模型分别对八种数字调制信号及三种模拟调制信号进行识别。按照控制变量原理设置了两组对比实验,将该文方法与现有方法进行对比,实验结果表明,该文方法在信噪比为-10~14 dB时,识别准确率提高了0%~9%,在信噪比为0 dB时准确率提高了近9%,由实验结果可知该文提出的方法优于现有方法。

Aiming at the problem of insufficient recognition accuracy of modulation signals,a recognition method based on convolutional neural network and attention mechanism is proposed. This method adds an attention mechanism between the convolution layer and the pooling layer of the convolution neural network,gives higher weight to the key characteristics of the modulation signal,carries out time-frequency analysis on the signal and converts it into a spectrum diagram as the input of the model. The model respectively analyzes eight kinds of digital modulation signals three kinds of analog modulation signals for identification. Two groups of comparative experiments are set up according to the principle of control variables. The experimental results show that the recognition accuracy of this method is improved by 0% to 9% at-10 dB to 14 dB signal-to-noise ratio and nearly 9% at 0 dB signal-to-noise ratio. The experimental results show that the method proposed in this paper is better than the existing methods.