对偶传播网络解决多元确知信号检测

A Solution of Detection of Certain Multi-signal Applied to Counter Propagation Network

针对利用匹配滤波器设计相关接收机可以实现二元确知信号检测,但对多元确知信号,该方法的门限值的选择与判决变得复杂的情况,将对偶传播网络(CPN)作为一种竞争型神经网络,通过对期望输出编码,实现多元确知信号的检测,较好地解决了门限值的选择与判决复杂的问题。由于CPN采用内星权和外星权的网络结构,仿真中采用信号的期望和方差作为神经网络的输入,收敛速率比采用采样点作为网络输入要快。仿真结果证实了该方法的有效性。

Correlation receiver is able to achieve the detection of ascertained dual signal applying to matched filter. However, to ascertained multi - signal, the selection and judgment of threshold quantity is awfully complicated. Counter propagation network （CPN）, which is a competitive neural network, can achieve the detection of ascertained multi - signal. Through the coding of expect output, CPN solves the problem of the selection and judges the threshold quantity well. Due to the structure of network synthesizes the algorithm of the instar weight vector and outstar weight vector, in our paper, expectation and variance of signal were used as the input of network, the computation speed was faster than that of sample value as input of network. Simulation results confirm the effectiveness of the method.