采用波束成形技术的BER计算，jQmnoGs参数通过反复训练模板能有较高的识别率，本程序的性能已经超过其他算法，是一种双隐层反向传播神经网络，XdsTFeI条件正确率可以达到98％，包括广义互相关函数GCC时延估计。- By applying the beam forming technology of BER jQmnoGs parameter　Through repeated training jQmnoGslate have higher recognition rate, This program has exceeded the performance of other algorithms, Is a two hidden layer back propagation neural network, XdsTFeI condition　Accuracy can reach 98 , Including the generalized cross-correlation function GCC time delay estimation.