摘要
目标噪声特征提取和目标分类器设计是被动声纳目标识别系统的关键技术.本文针对被动声纳目标识别,提出了一种新的调制连续谱特征提取方法.此外,为训练神经网络目标分类器,本文将遗传算法和BP算法相结合,提出了一种新的自适应遗传BP算法.最后,对海上实录的三类目标噪声进行了分类识别,实验结果表明本文设计的被动声纳目标识别系统具有很好的分类效果.
Feature extraction of targets radiated-noise and design of targets classifier are key techniques of passive sonar target recognition system. In this paper, a new feature extraction method of modulation spectrum is proposed first, and then a new adaptive genetic-backpropagation algorithm is proposed for training the neural network target classifier. Lastly, the classification experiment on three different classes of targets is done. Results of the experiment show that the passive sonar target recognition system designed in the paper has high correct classification rate.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
2002年第3期359-362,共4页
Pattern Recognition and Artificial Intelligence
关键词
被动声纳目标识别
调制连续谱
特征提取
自适应遗传BP算法
Passive Sonar Target Recognition, Modulation Spectrum, Feature Extraction, Adaptive Genetic- Backpropagation Algorithm