针对传统检测方法在复杂干扰条件下鲁棒性不足的问题,提出一种基于混合专家(Mixture-of-Experts,MoE)网络的抗干扰检测方法,用于复杂电磁环境下正交频分复用(Orthogonal Frequency Division Multiplexing,OFDM)信号的可靠检测。依托复...针对传统检测方法在复杂干扰条件下鲁棒性不足的问题,提出一种基于混合专家(Mixture-of-Experts,MoE)网络的抗干扰检测方法,用于复杂电磁环境下正交频分复用(Orthogonal Frequency Division Multiplexing,OFDM)信号的可靠检测。依托复杂干扰条件下的OFDM信号模型,采用融合导频信息的多通道接收特征作为模型输入;设计干扰识别驱动的级联接收结构,通过干扰识别网络判别干扰类型,并根据识别结果自适应选择MoE网络中的对应专家模型,实现针对性干扰抑制;结合DeepRx检测网络实现比特级检测。仿真结果表明,在多类典型干扰场景下,所提方法相比直接检测、混合干扰训练模型以及线性最小均方误差(Linear Minimum Mean Square Error,LMMSE)检测器,能够获得更低的误码率(Bit Error Rate,BER),并表现出更稳定的检测性能,从而有效提升复杂电磁环境下OFDM系统的抗干扰检测能力。展开更多
In the Digital World scenario,the confidentiality of information in video transmission plays an important role.Chaotic systems have been shown to be effective for video signal encryption.To improve video transmission ...In the Digital World scenario,the confidentiality of information in video transmission plays an important role.Chaotic systems have been shown to be effective for video signal encryption.To improve video transmission secrecy,compressive encryption method is proposed to accomplish compression and encryption based on fractional order hyper chaotic system that incorporates Compressive Sensing(CS),pixel level,bit level scrambling and nucleotide Sequences operations.The measurement matrix generates by the fractional order hyper chaotic system strengthens the efficiency of the encryption process.To avoid plain text attack,the CS measurement is scrambled to its pixel level,bit level scrambling decreases the similarity between the adjacent measurements and the nucleotide sequence operations are done on the scrambled bits,increasing the encryption.Two stages are comprised in the reconstruction technique,the first stage uses the intra-frame similarity and offers robust preliminary retrieval for each frame,and the second stage iteratively improves the efficiency of reconstruction by integrating inter frame Multi Hypothesis(MH)estimation and weighted residual sparsity modeling.In each iteration,the residual coefficient weights are modified using a mathematical approach based on the MH predictions,and the Split Bregman iteration algorithm is defined to resolve weighted l1 regularization.Experimental findings show that the proposed algorithm provides good compression of video coupled with an efficient encryption method that is resistant to multiple attacks.展开更多
摘要针对传统检测方法在复杂干扰条件下鲁棒性不足的问题,提出一种基于混合专家(Mixture-of-Experts,MoE)网络的抗干扰检测方法,用于复杂电磁环境下正交频分复用(Orthogonal Frequency Division Multiplexing,OFDM)信号的可靠检测。依托复杂干扰条件下的OFDM信号模型,采用融合导频信息的多通道接收特征作为模型输入;设计干扰识别驱动的级联接收结构,通过干扰识别网络判别干扰类型,并根据识别结果自适应选择MoE网络中的对应专家模型,实现针对性干扰抑制;结合DeepRx检测网络实现比特级检测。仿真结果表明,在多类典型干扰场景下,所提方法相比直接检测、混合干扰训练模型以及线性最小均方误差(Linear Minimum Mean Square Error,LMMSE)检测器,能够获得更低的误码率(Bit Error Rate,BER),并表现出更稳定的检测性能,从而有效提升复杂电磁环境下OFDM系统的抗干扰检测能力。
摘要In the Digital World scenario,the confidentiality of information in video transmission plays an important role.Chaotic systems have been shown to be effective for video signal encryption.To improve video transmission secrecy,compressive encryption method is proposed to accomplish compression and encryption based on fractional order hyper chaotic system that incorporates Compressive Sensing(CS),pixel level,bit level scrambling and nucleotide Sequences operations.The measurement matrix generates by the fractional order hyper chaotic system strengthens the efficiency of the encryption process.To avoid plain text attack,the CS measurement is scrambled to its pixel level,bit level scrambling decreases the similarity between the adjacent measurements and the nucleotide sequence operations are done on the scrambled bits,increasing the encryption.Two stages are comprised in the reconstruction technique,the first stage uses the intra-frame similarity and offers robust preliminary retrieval for each frame,and the second stage iteratively improves the efficiency of reconstruction by integrating inter frame Multi Hypothesis(MH)estimation and weighted residual sparsity modeling.In each iteration,the residual coefficient weights are modified using a mathematical approach based on the MH predictions,and the Split Bregman iteration algorithm is defined to resolve weighted l1 regularization.Experimental findings show that the proposed algorithm provides good compression of video coupled with an efficient encryption method that is resistant to multiple attacks.