【摘要】 在水声信号的研究中,应提高准确性。由于水声信号不具有严格的线性区域,过去直接对近似线性区域进行最小二乘拟合求得分形维数误差较大,不利于信号的识别。为解决上述问题,在分形布朗运动的基础上,提出采用水声目标信号近似线性区域进行了分段处理,运用最小二乘拟合求得各段的分形维数,并将各段的分形维数作为目标信号的特征矢量,进而以BP神经网络作为分类器进行目标的分类识别。实测目标的分类结果表明,上述方法提高了分形维数的稳定性,为水声目标信号的检测和识别提供良好的理论依据,并给出了具体的算法的步骤。更多还原

【Abstract】 In order to reduce the error when evaluating fractal dimensions of underwater acoustic targets,this paper proposes a method based on fractal Brown motion.First,the linear region of underwater acoustic targets is partitioned.Then,the least square fitting is used to compute the fractal dimensions of every region,which are the feature of the underwater acoustic targets.Furthermore,the recognition method based on back propagation network is proposed.The results of the recognition of actual targets show that the method can be used as a new feature of ship signals for detection and identification of ships.Finally,the algorithm is illustrated numerically in target recognition problem.更多还原