【摘要】 从目标的主动声呐回波中提取其特征信息是实现水下目标分类识别的有效手段。动物声呐在水下目标识别中表现出的优异性能为人工声呐提供了解决方法。文中以宽吻海豚喀啦信号作为主动声呐的发射信号,分别利用小波变换和海豚听觉系统模型2种方法提取了目标回波特征并作为支持向量机的输入进行分类。将通过海豚听觉系统模型得到的时谱图作为卷积神经网络的输入,对目标进行分类识别。研究表明,相比于小波变换方法,利用基于海豚听觉系统模型的特征提取方法进行目标分类识别的效果更好;结合卷积神经网络,采用海豚喀啦信号结合海豚听觉系统模型在水下目标识别中可以获得更好的结果。更多还原

【Abstract】 Extracting feature information from the active sonar echo of target is an effective method to realize underwater target classification and identification. The excellent performance of animal sonar in underwater target identification provides a solution for artificial sonar. In this study, the click signal of bottlenose dolphin is used as the transmitting signal of active sonar. The echo characteristics of the target are extracted by wavelet transform and model of dolphin auditory system, and are classified as the input of support vector machine. In addition, this paper proposes an idea that the time spectrum obtained by the model based on dolphin auditory system is used as the input of convolution neural network to classify and identify the target. The results show that compared with that of the wavelet transform method, the feature extraction method based on the computer model of dolphin auditory system is better for target classification and identification. Combining with convolution neural network, using dolphin click signal and model of dolphin auditory system can obtain better results in underwater target identification.更多还原