论文部分内容阅读
提出了一种用于船舶噪声目标分类的自适应高斯神经网络分类方法．首先利用傅里叶变换对三类船舶噪声进行预处理，然后利用高斯函数特性，将其和神经网络结合构成自适应高斯神经网络对目标信号谱进行有效识别特征自动提取和分类．该方法获得的特征空间与以AR建模和子带平均功率诸方法获得的特征空间相比，类别之间的可分性好，类间聚集性强。分类结果令人满意，证明了该方法的优越性．
An adaptive Gaussian neural network classification method for ship noise target classification is proposed. Firstly, three types of ship noise are preprocessed by using Fourier transform. Then, Gaussian function is used to combine with the neural network to form an adaptive Gaussian neural network to extract and classify the target signal spectrum effectively. Compared with the feature space obtained by the methods of AR modeling and subband average power, the feature space obtained by this method has good separability between classes and strong inter-class clustering. The classification result is satisfactory, which proves the superiority of this method.