基于深度学习的宽带信号检测技术

Wideband Signal Detection Technology Based on Deep Learning

提出一种将深度学习领域的目标检测模型应用于宽带密集信号检测问题的解决思路。将原始宽带采样数据流转化为时频图,将时频图作为目标检测网络的输入,通过目标检测网络的推断预测实现信号在时频图上的定位和识别,从而得到信号的起止时间、起止频率和类别信息。基于仿真生成的密集信号数据集,对MaskR-CNN模型进行训练和测试。实验结果表明,信噪比为0dB~14dB条件下,检测准确率为98%,召回率为93%,验证了基于深度学习的宽带信号检测算法的可行性。

7A new method to solve the problem of wideband dense signal detection is proposed,which applies the object detection model in the field of deep learning.Firstly,the original broadband sampling data stream is transformed into time frequency diagram,which is used as the input of the target detection network.The location and identification of the signal on the time frequency diagram are realized through the inference and prediction of the target detection network,so as to obtain the start and stop time,start and stop frequency and category information of the signal.Based on the dense signal data set generated by simulation,the Mask R-CNN model is trained and tested.The experimental results show that the detection accuracy is 98%and recall rate is 93%when the SNR is 0 dB~14 dB,which can verify the feasibility of wideband signal detection algorithm based on deep learning.