摘要
针对舰船目标的分类识别,研究了从解调制(DEMON)谱的线谱和连续线谱中提取特征向量的方法,提出了1种改进的对向传播网络(MCPN)分类模型及算法。通过对海上实录的3类水中目标辐射噪声进行分类识别实验,实验结果证明:MCPN的分类能力及对未训练目标的适应性优于传统的对向传播网络(CPN)和误差反向传播网络(BPN)。
A feature extraction method from line spectrum and continuous spectrum of DEMON is studied. The MCPN (Modified Counter Propagation Network) and classification algorithm are put forward. The classification recognition experimental results for three different classes of target noises in the sea show that MCPN has higher correct recognition rate and better adaptability to untrained targets than CPN(Counter Propagation Network) or BPN (Back Propagation Network).
出处
《电声技术》
2007年第8期4-6,20,共4页
Audio Engineering
关键词
特征提取
神经网络
目标分类
feature extraction
neural network
target classification