Application Of Joint Time-Frequency Target In Sea-Clutter .

3y ago
39 Views
2 Downloads
1.77 MB
49 Pages
Last View : 14d ago
Last Download : 3m ago
Upload by : Kian Swinton
Transcription

Defence Research andDevelopment CanadaRecherche et developpementpour la d6fense CanadaDEFENCE7DEFENSEApplication of Joint Time-FrequencyRepresentations to a Maneuvering AirTarget in Sea-Clutter: Analysis Beyond FFTT. Thayaparan and S. KennedyDISTRIBUTION STATEMENT AApproved for Public ReleaseDistribution UnlimitedDefence R&D Canada - OttawaTECHNICAL MEMORANDUMDRDC Ottawa TM 2003-090March 2003Canada20030910 169

Application of Joint Time-FrequencyRepresentations to a Maneuvering Air Targetin sea-clutter: analysis beyond FFTT ThayaparanDefence R&D Canada - OttawaS. KennedyCarleton UniversityDefence R&D Canada - OttawaTechnical MemorandomDRDC Ottawa TM 2003-090March 2003

@ Her Majesty the Queen as represented by the Minister of National Defence, 2003@ Sa majest6 la reine, repr sent e par le ministre de la Defense nationale, 2003miV

AbstractTraditionally, radar signals have been analysed in either the time or the frequencydomain. Joint time-frequency representations characterize signals over atime-frequency plane. They thus combine time-domain and frequency-domain analysesto yield a potentially more revealing picture of the temporal localization of a signal'sspectral components. Therefore, for air target returns with time-varying frequencycontent, the joint time-frequency representations offer a powerful analysis tool. Aconcise review of time-frequency transforms is provided as background and is neededto appreciate how time-frequency processing methods can improve conventional timeor frequency processing methods. The report then describes and illustrates theadvantages of using joint time-frequency techniques to analyze a multi-componentsignal, a noisy signal, and experimental aircraft data. Finally, we use time-frequencyanalysis techniques for the detection of maneuvering aircraft using HF radar in heavilycluttered regions. We compare the ability of different time-frequency transforms toresolve several experimental aircraft returns. The results clearly demonstrate thattime-frequency analysis techniques can significantly improve the detectionperformance of the HF radar and add considerable physical insight over what can beachieved by conventional Fourier transform methods currently used by HF radars.DRDC Ottawa TM 2003-090

Resume'Par le pass6, les signaux radar ont W analyses soit dans le domaine temporel, soit dansle domaine fr quentiel. Les representations temps-fr quence mixtes montrent lescaract ristiques des signaux sur un plan temps-fr quence. Elles combinent ainsi lesanalyses du domaine temporel et du domaine fr quentiel pour donner une imagepotentiellement plus r vlatrice de l'emplacement temporel des composantes spectralesd'un signal. Par consequent, les representations temps-fr quence mixtes constituent unoutil d'analyse puissant des signaux dont le contenu fr quentiel varie dans le temps.Nous pr sentons une breve 6tude des transformations temps-fr quence afin de fournir labase n cessaire pour appr cier la fagon dont les m thodes de traitementtemps-fr quence pennettent d'am liorer les m thodes classiques de traitement temporelou fr quentiel. Ensuite, le rapport d crit et illustre les avantages de l'utilisation destechniques temps-fr quence mixtes pour l'analyse d'un signal Aplusieurs composantes,d'un signal perturb6 par le bruit et de donn es exp rimentales d'a ronef. Enfin, nousutilisons les techniques d'analyse temps-fr6quence pour la detection d'un a ronefeffectuant des manceuvres en ayant recours au radar HF dans des zones Ainiveau declutter 6lev6. Nous comparons la capacit6 de diffhrentes transfornationstemps-fr quence pour r soudre plusieurs signaux exp rimentaux d'adronef. Lesr sultats montrent clairement que les techniques d'analyse temps-fr quence pennettentd'am 1iorer de fagon appreciable l'efficacit6 de detection du radar HF et d'acqu rirbeaucoup d'autres connaissances physiques par comparaison aux m thodes Aitransformation de Fourier classiques qui sont actuellement utilis es par les radars HF.iiDRDC Ottawa TM 2003-090

Executive summaryOne of the central problems in exploiting High Frequency (HF) radar data is theanalysis of a time series. The problem at hand is how to extract the information presentin the data and use it to its full potential. In order to address this problem we turn to thefield of signal analysis and data representations. Traditionally, radar signals have beenanalysed in either the time or the frequency domain. The Fourier Transform is at theheart of a wide range of techniques that are used in HF radar data analysis andprocessing. Mapping the data into the temporal frequency domain is an effective way ofrecording the data such that their global characteristics can be assessed. However, thechange of frequency content with time is one of the main features we observe in HFradar data. Because of this change of frequency content with time, radar signals belongto the class of non-stationary signals. The analysis of non-stationary signals requirestechnique that extend the notion of a global frequency spectrum to a local frequencydescription. The spectral energy density function that is obtained by means of a FourierTransform, the so-called power spectrum, shows the frequencies that are present in ourdata, but does not reveal where changes in the frequency content occur. Consequently,for the interpretation of radar data in terms of a changing frequency content, we need arepresentation of our data as a function of both time and frequency. Only, quiterecently, the joint time-frequency representation of signals has become a major area ofresearch in radar signal processing.The time-frequency representations characterize signals over a time-frequency plane.They thus combine time-domain and frequency-domain analyses to yield a potentiallymore revealing picture of the temporal localization of a signal's spectral components.They may also serve as a basis for signal detection, imaging, characterization, coding,and processing. A complete and comprehensive theory for joint time-frequencyanalysis does not yet exist. There is no unique time-frequency representation of a signalthat satisfies all the properties of a physically correct joint time-frequency energydensity function. However, discarding the requirement that all properties must besatisfied in one time-frequency representation, a class ofjoint time-frequencyrepresentations can be derived that serves as a model of a local power spectrum.In this report, a concise review of time-frequency transforms is provided as backgroundand is needed to appreciate how time-frequency processing methods can improveconventional time or frequency processing methods. The report then describes andillustrates the advantages of using joint time-frequency techniques to analyze amulti-component signal, a noisy signal, and experimental aircraft data. Finally, we usetime-frequency analysis techniques for the detection of maneuvering aircraft using HFradar in heavily cluttered regions. We compare the ability of different time-frequencytransforms to resolve several experimental aircraft signals. The results clearlydemonstrate that time-frequency analysis techniques can significantly improve thedetection performance of the HF radar and add considerable physical insight over whatcan be achieved by conventional Fourier transform methods currently used by HFradars.DRDC Ottawa TM 2003-090iii

T. Thayaparan, S. Kennedy. 2003. Application of Joint Time-Frequency Representations to aManeuvering Air Target in sea-clutter: analysis beyond FFT. DRDC Ottawa TM 2003-090.Defence R&D Canada - Ottawa.ivDRDC Ottawa TM 2003-090

SommaireL'analyse d'une s rie chronologique constitue P'un des principaux probl mes 6prouv savec les donndes du radar haute fr quence (BF). Le prob1 me qui se pose consiste dtrouver un moyen pour extraire l'information contenue dans les donn es et pourl'utiliser Ason plein potentiel. Pour r soudre ce probkme, nous faisons appel audomaine de l'analyse des signaux et de la representation des donn es. Par le pass6, lessignaux radar ont W analys s soit dans le domaine temporel, soit dans le domainefr quentiel. La transformation de Fourier est au cceur d'une vaste gamme de techniquesqui sont utilis es pour l'analyse et le traitement des donn es du radar BY. La mise encorrespondance des donn es dans le domaine des fr quences temporelles constitue unmoyen efficace pour enregistrer les donn es de telle fagon que leurs caract ristiquesglobales puissent 8tre 6valu es. Cependant, la variation du contenu fr quentiel avec letemps est une des principales caract ristiques que nous observons dans les donn es duradar BE.En raison de cette variation du contenu fr quentiel avec le temps, les signauxradar appartiennent Aila cat gorie des signaux non stationnaires. Pour l'analyse dessignaux non stationnaires, ii faut avoir recours Ades techniques qui 6largissent la notiond'un spectre de fr quences global Aune description de fr quence locale. La fonction dedensit6 d'6nergie spectrale obtenue au moyen d'une transformation de Fourier, etappel e le spectre de puissance, montre quelles fr quences sont contenues dans nosdonn es, mais n'indique pas ofi se produisent les changements du contenu fr quentiel.Par consequent, pour l'interpr tation des donn es radar du point de vue d'un contenufr quentiel changeant, nous avons besoin d'une reprdsentation de nos donn es enfonction du temps et de la fr quence. Ce W'est que tout r cemment que in representationtemps-fr quence mixte des signaux est devenue un sujet important de recherche dans letraitement des signaux radar.Les repr6sentations temps-fr quence mixtes montrent les caract ristiques des signauxsur un plan temps-frdquence. Elles combinent ainsi les analyses du domaine temporelet du domaine fr quentiel pour donner une image potentiellement plus r vdlatrice del'emplacement temporel des composantes spectrales d'un signal. Elles peuvent aussiservir de base pour la detection des signaux, l'imagerie, la caract risation, le codage etle traitement. 11 n'existe pas encore de th orie compl te et d taill e pour l'analysetemps-fr quence mixte. 11 n'existe pas de representation temps-fr quence unique d'unsignal qui satisfait Atoutes les propridt s d'une fonction de densit6 d'6nergietemps-fr quence mixte correcte du point de vue physique. Cependant, si on faitabstraction de l'exigence selon laquelle ii faut satisfaire d toutes les propri t s dans uneseule representation temps-fr quence, on peut dtablir une classe de representationstemps-fr quence mixtes qui sert de mod le pour un spectre de puissance local.Dans ce rapport, nous pr sentons une breve dtude des transformations temps-fr quenceafin de foumir la base n cessaire pour appr cier la fagon dont les m thodes detraitement temps-fr quence permettent d'am liorer les m thodes classiques detraitement temporel ou fr quentiel. Ensuite, le rapport d crit et illustre les avantages del'utilisation des techniques temps-fr quence mixtes pour l'analyse d'un signal ADRDC Ottawa TM 2003-090v

plusieurs composantes, d'un signal perturb6 par le bruit et de, donn es exp rimentalesd'a ronef. Enfin, nous utilisons les techniques d'analyse temps-frdquence pour ladetection d'un a ronef effectuant des manceuvres en ayant recours au radar HF dans deszones Aniveau de clutter 61ev6. Nous comparons la capacit6 de diff rentestransformations temps-fr quence pour r soudre plusieurs signaux exp rimentauxd'a ronef. Les r sultats montrent clairement que les techniques d'analysetemps-fr quence permeftent d'amnd1iorer de fagon apprdciable 1'efficacit6 de detectiondu radar HF et d'acqu rir beaucoup d'autres connaissances physiques par comparaisonaux m thodes Atransformation de Fourier classiques qui sont actuellement utilis es parles radars HF.T. Thayaparan , S. Kennedy. 2003. Application of Joint Time-Frequency Representations to aManeuvering Air Target in sea-clutter: analysis beyond FFT. DRDC Ottawa TM 2003-090. R& D pour la defense Canada - Ottawa.viDRDC Ottawa TM 2003-090

Table of contentsAbstract .Rdsumn.i. .iiExecutive summary .iiiSommaire. .Table of contents .List of figures .v.vii.ix1.Introduction .12.Advantages of Time-Frequency Representations .32.1Detecting Multiple Signals .32.2Extracting a Signal from Noise .52.3Signal Analysis .83.Comparison of Time-Frequency Distributions .113.1Gabor Representation .123.2Margenau-Hill Spectrogram .123.3Born-Jordan Distribution .123.4Binomial Distribution .133.5Choi-Williams Distribution .133.6Smoothed Wigner-Ville Distribution .133.7Spectrogram Distribution .143.8Zhao-Atlas-Marks Distribution .143.9Reassigned Gabor Representation .143.10Reassigned Morlet Scalogram .153.11Reassigned Spectrogram Distribution .163.12Adaptive Energy Distribution .16DRDC Ottawa TM 2003-090.Vii

4.5.Comparison of Transforms .184.1Comparison Signals .184.2Constant Frequency .194.3Non-Uniform Chirp near Clutter .204.4Chirp Through Clutter .204.5Changing Signal .214.6Small Frequency Modulation .21Evaluation of Transforms .285.1Gabor Representation .285.2Margenau-Hill Spectrogram .285.3Born-Jordan Distribution .285.4Binomial Distribution .285.5Choi-Williams Distribution .285.6Smoothed Wigner-Ville Distribution .295.7Spectrogram Distribution .295.8Zhao-Atlas-Marks Distribution .295.9Reassigned Gabor Representation .295.10Reassigned Morlet Scalogram .295.11Reassigned Spectrogram Distribution .305.12Adaptive Energy Distribution .306.Overall Evaluation .317.Conclusion .32References .viii33DRDC Ottawa TM 2003-090

List of figures1Time series of multi-component signal .32The spectrum of the multi-component signal .43The Margenau-Hill spectrogram of the multi-component signal .44Time series of a signal in white Gaussian noise .55The spectrum of a signal in white Gaussian noise .66A time-frequency representation of a signal in white Gaussian noise .67Filtered time-frequency transform and time signals .78Spectrum of noisy aircraft data .99The Gabor representation of the aircraft data .910Filtered Gabor representation .1011Recovered target signal .1012Spectrums of the five comparison signals .1813Time-frequency representations of a signal with constant frequency .14Time-frequency representations of a signal near clutter .2415Time-frequency representations of a signal through clutter .2516Time-frequency representations of the changing signal .2617Time-frequency representation of a signal with small frequency modulation .27DRDC Ottawa TM 2003-090.23ix

This page intentionally left blank.xDRDC Ottawa TM 2003-090

1.IntroductionOne fundamental mathematical tool employed in radar signal processing is a transform.When asked to multiply the Roman numerals LXIV and XXXII, only a few of us willbe able to give the correct answer right away. However, if the Roman numerals are firsttranslated into Arabic numerals, 64 and 32, then all of us can get 2048 immediately.The process of converting the unfamiliar Roman numerals into common Arabicnumerals is a typical example of transforms [1]. By properly applying transforms, wecan simplify calculations or make certain attributes of the signal explicit.One of the most popular transforms known to scientists and engineers is the Fouriertransform that converts a signal from the time-domain to the frequency-domain. In fact,the Fourier transform is not simply a mathematical tool to make calculations easier. Italso acts as a mathematical prism to break down a signal into a group of waveforms(different frequencies), as a prism breaks up light into a color spectrum. The Fouriertransform is so powerful that people tend to apply it everywhere without noticing onefundamental difference between the mathematical prism and a real prism. Thespectrum produced by the prism in the morning is different from that in the evening.We may say that the prism gives instantaneous spectra. Using a prism to examinespectra of light, there is no need for the information about light that existed millionyears ago and the light that will be there tomorrow. However, this is not the case forFourier transform. To compute the Fourier transform, we not only need previousinformation, but also information that have not yet occurred. The spectrum computedby the Fourier transform is the spectrum averaged over an infinitely long time beforethe present to an infinitely long time after the present! [2]The time series signals of radar targets have been traditionally analyzed in thefrequency-domain. The transformation of a time series into a frequency-domain usingthe Fourier transform provides useful information regarding the target. However,Fourier spectral analysis is not without its shortcomings. The most prominentshortcoming of Fourier spectral analysis is its inability to properly represent frequencycontents that changes over an observation period. 'The most widely used signalprocessing tool is the FFT (fast Fouriertransform); the most widely misused signalprocessing tool is also the FFT" [2]. Fourier transform-based techniques are effectiveas long as the frequency contents of the signal do not change with time. However, whenthe frequency contents of the data samples evolve over an observation period, otheralternative methods should be considered. Joint time-frequency representations weredeveloped to overcome this limitation of Fourier spectral analysis [3-4].The joint time-frequency analysis has been a topic of much interest in the signalprocessing community in the past decade. Over the past ten years, radar researchershave also investigated time-frequency transforms as a unique tool for radar-specificsignal analysis and image processing applications [2,5-9]. Both traditionaltime-frequency techniques, as well as the new tools developed in the signal processingcommunity, have been applied to various radar problems. Like the developments inDRDC Ottawa TM 2003-090

other fields, such as underwater acoustics and speech processing, it was found thattime-frequency transforms provide additional insight into the analysis, interpretation,and processing of radar signals that is sometimes superior to what is achievable in thetraditional time or freque

le traitement. 11 n'existe pas encore de th orie compl te et d taill e pour l'analyse temps-fr quence mixte. 11 n'existe pas de representation temps-fr quence unique d'un signal qui satisfait A toutes les propridt s d'une fonction de densit6 d'6nergie temps-fr quence mixte correcte du point de vue physique. Cependant, si on fait

Related Documents:

Weasler Aftmkt. Weasler APC/Wesco Chainbelt G&G Neapco Rockwell Spicer Cross & Brg U-Joint U-Joint U-Joint U-Joint U-Joint U-Joint U-Joint U-Joint Kit Stock # Series Series Series Series Series Series Series Series 200-0100 1FR 200-0300 3DR 200-0600 6 L6W/6RW 6N

REFERENCE SECTION NORTH AMERICAN COMPONENTS John Deere John Deere Aftmkt. John Deere APC/Wesco Chainbelt G&G Neapco Rockwell Spicer Cross & Brg U-Joint U-Joint U-Joint U-Joint U-Joint U-Joint U-Joint U-Joint Kit Stock # Series Series Series Series Series Series Series Series PM200-0100 1FR PM200-0300 3DR

Bones and Joints of Upper Limb Regions Bones Joints Shoulder Girdle Clavicle Scapula Sternoclavicular Joint Acromioclavicular Joint Bones of Arm Humerus Upper End: Glenohumeral Joint Lower End: See below Bones of Forearm Radius Ulna Humeroradial Joint Humeroulnar Joint Proximal Radioulnar Joint Distal Radioulnar Joint Bones of Wrist and Hand 8 .File Size: 2MBPage Count: 51

Procedure Code Service/Category 15824 Neurology 15826 Neurology 19316 Select Outpatient Procedures 19318 Select Outpatient Procedures 20930 Joint, Spine Surgery 20931 Joint, Spine Surgery 20936 Joint, Spine Surgery 20937 Joint, Spine Surgery 20938 Joint, Spine Surgery 20974 Joint, Spine Surgery 20975 Joint, Spine Surgery

Procedure Code Service/Category 15824 Neurology 15826 Neurology 19316 Select Outpatient Procedures 19318 Select Outpatient Procedures 20930 Joint, Spine Surgery 20931 Joint, Spine Surgery 20936 Joint, Spine Surgery 20937 Joint, Spine Surgery 20938 Joint, Spine Surgery 20974 Joint, Spine Surgery 20975 Joint, Spine Surgery

Question of whether density really improved Also constructability questions INDOT Joint Specification Joint Density specification? More core holes (at the joint!) Taking cores directly over the joint problematic What Gmm to use? Joint isn't vertical Another pay factor INDOT Joint Specification Joint Adhesive Hot applied

2 Mirror Frequency Filter 2.1 Mirror Frequency In radio reception using heterodyning in the tuning process, the mirror frequency is an undesired input frequency that is capable of producing the same intermediate frequency (IF) that the desired input frequency produces. It is a potential source of interference to proper reception.

Joint Publication (JP) 3-0 is the keystone document in the joint operations series and is a companion to joint doctrine’s capstone JP 1, Doctrine for the Armed Forces of the United States. It provides guidance to joint force commanders (JFCs) and their subordinates to plan, execute, and assess joint military operations.