摘要
为在舰船目标识别、分类中有效选取特征量,采用小波变换获取舰船辐射噪声信号的小波能量谱,然后计算不同类目标信号的类平均小波能量谱之间距离信号的λ水平能量聚点,据此选取能量谱差异显著的频段,最后对各能量谱作聚类分析实现特征压缩并确定特征量。在舰船声场通过特性研究中的应用表明,该方法具有识别与分类准确率高的特点。
Based on wavelet power-spectrum, the modeling method for feature of ship radiated-noise was studied. The wavelet power-spectrum of ship radiated-noise was obtained by discrete wavelet transform. Power condensation point on λ level of signal was defined. On the base, both distance signals between the wavelet power-spectrum of different kind of ship radiated-noise and their power condensation point were calculated. Feature frequency bands were selected according to the level of power condensation point, and power at feature frequency bands was used as feature vector. It is shown that the method has high accuracy for target recognition and classification by application in ship acoustic through signature.
出处
《系统仿真学报》
EI
CAS
CSCD
北大核心
2007年第17期4025-4027,共3页
Journal of System Simulation
基金
海军工程大学自然科学基金(HGDJJ07006)
关键词
小波能量谱
λ水平能量聚点
分类
特征量
wavelet power-spectrum
power condensation point on 2 level
recognition and classification
feature vector
