哈尔滨工程大学学报
哈爾濱工程大學學報
합이빈공정대학학보
JOURNAL OF HARBIN ENGINEERING UNIVERSITY
2013年
8期
967-971
,共5页
水下目标%特征提取%经验模态分解%WVD方法
水下目標%特徵提取%經驗模態分解%WVD方法
수하목표%특정제취%경험모태분해%WVD방법
underwater target%feature extraction%empirical mode decomposition%WVD method
针对传统时频分析方法在水下目标特征提取时的局限性,采用一种基于经验模态分解和 Wigner-Ville 分布( Wigner-Ville distribution,WVD)的提取水下小目标亮点特征的新方法。该方法可以将多分量信号分解为一组单分量信号进行分析,不仅可以有效地抑制WVD分布中多分量信号所产生的严重的交叉项干扰,而且解决了噪声对HHT谱产生波动的影响,通过仿真分析和实验数据的处理验证了该方法的有效性。对三类目标模型的回波进行了特征提取,并分析了各类模型回波的不同特点。其中,规则雷的回波以镜反射波为主;不规则雷的亮点结构相互叠加致使频带范围变宽;假雷的回波由于是多个信号的非线性叠加,使得到的时频特性分布较广、比较散乱。
針對傳統時頻分析方法在水下目標特徵提取時的跼限性,採用一種基于經驗模態分解和 Wigner-Ville 分佈( Wigner-Ville distribution,WVD)的提取水下小目標亮點特徵的新方法。該方法可以將多分量信號分解為一組單分量信號進行分析,不僅可以有效地抑製WVD分佈中多分量信號所產生的嚴重的交扠項榦擾,而且解決瞭譟聲對HHT譜產生波動的影響,通過倣真分析和實驗數據的處理驗證瞭該方法的有效性。對三類目標模型的迴波進行瞭特徵提取,併分析瞭各類模型迴波的不同特點。其中,規則雷的迴波以鏡反射波為主;不規則雷的亮點結構相互疊加緻使頻帶範圍變寬;假雷的迴波由于是多箇信號的非線性疊加,使得到的時頻特性分佈較廣、比較散亂。
침대전통시빈분석방법재수하목표특정제취시적국한성,채용일충기우경험모태분해화 Wigner-Ville 분포( Wigner-Ville distribution,WVD)적제취수하소목표량점특정적신방법。해방법가이장다분량신호분해위일조단분량신호진행분석,불부가이유효지억제WVD분포중다분량신호소산생적엄중적교차항간우,이차해결료조성대HHT보산생파동적영향,통과방진분석화실험수거적처리험증료해방법적유효성。대삼류목표모형적회파진행료특정제취,병분석료각류모형회파적불동특점。기중,규칙뢰적회파이경반사파위주;불규칙뢰적량점결구상호첩가치사빈대범위변관;가뢰적회파유우시다개신호적비선성첩가,사득도적시빈특성분포교엄、비교산란。
To deal with the limitations in feature extraction of an underwater target based on traditional time-frequen-cy analytical methods, a new method based on empirical mode decomposition and Wigner-Ville distribution ( WVD) is adopted to extract the highlight features of small underwater targets. Multiple-component signal can be decomposed into a collection of single component signals, which can not only restrain the severe cross-term interfer-ence in WVD method caused by multiple-component signals, but also solve the problem that HHT spectrum can be influenced by noise. Simulation analysis and experiment results have affirmed the effectivity of the method. Echo waves of three different target models are analyzed by feature extraction, and each gets a different conclusion. Echo waves of the regular mine models are mainly composed of specular reflection. Echo waves'superposition of the irreg-ular mine models widens the frequency band. And echo waves of false target are caused by several signals'nonlinear superposition, which widen the time-frequency feature distribution and make the highlight extraction harder.