基于时空神经网络增强数字示波器功能的研究

运用时空神经网络时域和空域模式识别方法给数字示波器增加AM调制信号测量功能。选择Elman神经网络结构,采用反向传播网络训练函数traingdx和learnbcf函数的算法,实现了AM调制信号检波的功能。同时,增加输出反馈回隐层的连接和延迟,采用附加动量因子的梯度下降权值/阀值学习算法改进神经网络。改进的网络学习速度快,逼近精度高,输出既没有振荡,也不产生纹波;并且网络适应性好,测量的鲁棒性高,要求采集信号周期少;方法新颖,运算量小,计算误差小,添加到数字示波器函数中,实现了示波器测量AM调制信号的功能。

Study of Enhancing Features of Digital Oscilloscope Based on Elman Networks

In this paper, a method for enhancing the measure performance of digital oscilloscope for Amplitude Modulation (AM) signals is presented by applying Elman sptio-temporal neural network. In this method, the demodulation of AM signals is implemented by adopting both"traingdx"and"learnbcf"functions in Elman network;the structure of Elman network is improved by introducing weights and delays from the output layer to the hidden layer;and an additive momentum factor is adopted in gradient learning algorithm. Simulation results demonstrate that the proposed method has faster learning speed, less computational error, and higher measuring robustness and precision.