摘要
被动声纳捕捉的目标一般是船舶或鱼雷的辐射噪声。该辐射噪声可视为周期性激励与信道传递函数的卷积构成。利用同态分析进行目标分类,即通过倒谱分析将各噪声分量变为线性相加关系,实现各信号分量以及信道的分离,以便提取信号特征进行分类。通过对实际信号的分析,证明水声信号的倒谱特征在概率意义上含有比较稳定的类别信息,基于支持向量机的分类实验结果,说明了根据倒谱特征识别被动水声目标的可行性。
Targets captured by passive sonar are currently the radiant noises of ships or torpedoes. These noises can be considered to he composed through convolution of many kinds of excitation source and the transform function of oceanic acoustic channel. A homomorphic analysis based classification method is proposed. The relationships between components of noise are converted to linear summarization through cepstrum transform. Then the components of noise and oceanic acoustic channel are separated. These processes are helpful to distill the signal features and to classify the targets. Through analyses of practical signals, it is proved that the cepstrum features of underwater acoustic signals contain relatively steady class information in a sense of probability. The experimental classification results show the feasibility of classifying passive underwater targets based on cepstrum features.
出处
《系统工程与电子技术》
EI
CSCD
北大核心
2005年第10期1708-1710,共3页
Systems Engineering and Electronics
基金
中国科学院王宽诚博士后工作奖励基金课题(20021025123820)
关键词
倒谱
被动声纳
目标分类
支持向量机
cepstrum
passive sonar
target classification
support vector machine