基于实虚型连续多值复数Hopfield神经网络的QAM盲检测

Blind Detection of QAM Signals with a Complex Hopfield Neural Network with Real-Imaginary-Type Soft-Multistate-Activation-Function

针对统计量算法盲检测QAM信号的缺陷,该文提出了一个实虚型连续多值复数Hopfield神经网络算法,该网络的实部、虚部各含一个连续多值实激活函数.该文构造了适用于该网络的能量函数,并分别在异步和同步更新模式下证明了该神经网的稳定性.当该神经网的权矩阵借助接收数据补投影算子构成时,该实虚型连续多值复数Hopfield神经网络可有效地实现QAM信号盲检测.仿真试验表明:该算法采用较短接收数据即可到达全局真解点,并且适用于含公零点信道.

Considering the disadvantage of the algorithms based on statistics, a novel algorithm based on Complex Hotrfield Neural Network with Real-Imaginary-type Soft-Multistate-activalion-function （CHNN_ RISM） is proposed to detect QAM signals blindly. A multi-valued continuous activation function is comtructed in both of the real part and imaginary part of CHNN_ RISM. A new energy function for CHON_ RISM is constructed in this paper and the stabilities with asynchronous and synchronous operating mode are also analyzed separately. While the weighted matrix of CHNN_ RISM is constructed by the complementary projection operator of received signals,the problem of quadratic optimization with integer consaaints can successfully solved with the CHNN_ RISM, and the QAM signals are blindly detected.Simulation results show that the algorithm reaches the real equilibrium points with shorter received signals and appropriate for channel with common zeros.