Abstract: A blind beamforming algorithm based higher-order cumulant and neural network is presented according to the characteristic of underwater acoustic environment and signal. This method combines the characteristic that higher-order cumulant can eliminate effect of Gaussian noise with structural advantage that neural network can operate in parallel. It can not only avoid computing the inverse of a matrix but also use easily hardware to carry out the blind optimizing beamforming. Simulation proves correctness of this algorithm.