高阶QAM信号的前馈神经网络相位修正水声信道盲均衡算法

提出了适用于复值信号的前馈神经网络盲均衡算法，并针对传统常模盲均衡算法不具备相位修正能力的缺点，对代价函数进行了改进，提出了基于前馈神经网络的修正常模盲均衡算法，同时针对算法起伏性大、收敛速度慢的问题，利用判决正方形方法进行了改进．仿真结果表明，在高阶QAM通信系统中，新的神经网络盲均衡算法不仅能够进行相位偏差的自修正，同时在算法的收敛能力、收敛速度以及稳健性方面都较传统神经网络常模算法更有优势．

Phase Self-amending Blind Equalization Algorithm Using Feedforward Neural Network for High-order QAM Signals in Underwater Acoustic Channels

A complex-valued blind equalization algorithm using feedforward neural network is brought for- ward. Aiming at the defects that traditional constant modulus equalization algorithm can＇t rectify the phase de- flection, the cost function is reformed and also a new modified constant modulus algorithm is given. Besides, the new algorithm is improved by introducing the square decision technique to achieve better convergence speed and less gurgitation. The results of simulation show that this new equalization algorithm not only has the ability of phase self-amending, but also performs better than traditional algorithm in the ability and speed of convergence in high order QAM communication systems.