Transient Interference Excision in China Experimental OTHR

Publication REE REE 2005-3
DOI : You do not have permission to access embedded form.


Transient Interference Excision in China Experimental OTHR


1.89 Mo


Creative Commons Aucune (Tous droits réservés)
<resource  xmlns:xsi=""
        <identifier identifierType="DOI">10.23723/1301:2005-3/20553</identifier><creators><creator><creatorName>Tieping Wu</creatorName></creator><creator><creatorName>Mengdao Xing</creatorName></creator><creator><creatorName>Sunjun Wu</creatorName></creator><creator><creatorName>Zheng Bao</creatorName></creator></creators><titles>
            <title>Transient Interference Excision in China Experimental OTHR</title></titles>
        <resourceType resourceTypeGeneral="Text">Text</resourceType><dates>
	    <date dateType="Created">Sat 21 Oct 2017</date>
	    <date dateType="Updated">Sat 21 Oct 2017</date>
            <date dateType="Submitted">Mon 19 Feb 2018</date>
	    <alternateIdentifier alternateIdentifierType="bitstream">89c67c63cdb9ee85176e0961b0562115a0c86f33</alternateIdentifier>
            <description descriptionType="Abstract"></description>

LES RADARS m m Transient Interference Excision m a in China Experimental OTHR ParTieping WU, Mengdao XING, Sunjun WU, Zheng BAO Key Laboratory for Radar Signal Processing, Xidian University China Mots clés Over theHorizon Radar (OTHR), Transient interference, Lightning impulsive, Meteortrailechoes 1. Introduction The skywave of over-the-horizon radar (OTHR) looks down at its targets from the ionosphere, so it has a large operating range and a large cover areas, but it received a large multitude of backscatterer echo form the terrain and sea, and strong interference, such as, radio frequency interference, industrial interference, impulsive noise, lightning impulsive, metcor trail echoes and so on, because it works on HF band. To detect the targets in strong clutter and interference, the resolution cell must be decrease. The resolution cell is three dimensional (3-D), that is range, azimuth, Doppler. The range resolution depends on the bandwidth of the transmit signal, but widening the bandwidth is restrained, because it works on frequency gap where external noise is weak, the pro- bability to appear wide frequency interval is small. The range resolution can be improved with desultory wide- band, but if the band has gap, the side-lobe will raise. The improvement of the azimuth resolution depends on the increase of antenna aperture. Now the length of OTHR antenna is the order of kilometer, so it is difficulty to increase. As a result the increase of the resolution cell depends on the improve of the Doppler resolution, while the Doppler resolution depends on coherent integration time, at the same time it is demanded that the echoes be strongly coherent in the coherent integration time, which requires the emitter and the receiver be of high stabiliza- tion (according to modern technology level, that is easy to meet). But for OTHR, the influence of phase perturbation and transient interference is great. The phase perturbation caused by electromagnetic wave propagation can produce significant clutter and target Doppler spreading, the phase perturbation correction is very similar to auto- focus in radar imaging, such as PGA and PPP methods [1-3]. Transient interference is the interference whose existing time is much short than the coherent integration time. This kind of interference has a very wide Doppler spectral band, which will influent almost all or large part c ESSENTIEL SYNOPSIS Le radar transhorizonà onde de ciel recherche les cibles après réflexion sur l'ionosphère. II est doté d'une grande portée et couvre une largezone, mais il reçoit de forts échos du terrain et de la mer,et il est soumisà de fortes interférences : interférences radiofréquences, bruit industriel, bruit impulsif, échos sur les trainées météoriques, etc. Les interférences peuvent être divi- sées en deuxclasses: longterme et transitoire. Les interférences transitoires sont courtes, mais de forte intensité. Dans cette communication,on commence paranalyser lescarac- téristiques des interférences,puis, en situation de fouillis de mer, on utilise unedécompositionen éléments proprespour séparerle fouillis de mer et le filtrer, ou bien,en situation de fouillis de terre, on filtre le fouillis de terre dans le domaine fréquentiel puis on revient dans le domaine temporel ; ensuite l'interférence transi- toire est détectée, puis excisée des échos reçus et finalement, les échos de cibles et de fouillis excisés sont prédits par l'algo- rithme de prédiction linéaire de Burg. Cette méthode de traite- ment a été appliquéeavecsuccèsà des données réelles du radar expérimentaltranshorizonchinois. High-frequency(HF) skywave in over-the-horizonradar (OTHR) looks down at its targets from the ionosphere,so it has a large operating rangeand a large cover areas, but it receivesa large- multitude of backscattererecho form the terrains and the seas, and strong interference, such as, radio frequency interference, industrial interference, impulsive noise, lightning impulsive, meteor trail echoes and so on. Interference can be divided into long-timedinterferenceandtransient interference,transient inter- ference lastsa shorttime, but its intensity isgreat.This paperfirst discussed the character of the transient interference, and used eigen-decompositionto separatethe seaclutter subspaceandfil- ter it, or filtered terrestrial clutterin frequencydomainand backto time domain,then the transient interference was detected, after that the transient interference was excised from the original echoes. Finally, the excised clutter and target echoes were pre- dicted by Burg linear prediction algorithm. This processing method hasbeen successfully appliedto the realdatafrom China experimentalOTHR. REE Nn 3 Mars 2005 M De G s s 9, 0 r LES RADARS of the Doppler ceil, thus it should be eliminated in the time-domain processing. The transient interference detection and excision problem is discussed in [4]. In the paper we use eigen-decomposition method to obtain the principal components that are clutters, then the transient interference is detected, after that the transient interferen- ce is excised from the original echoes, finally, the excised clutter and target echoes are predicted by using Burg linear prediction algorithm. This processing method has been successfully applied to the real data from China experimental OTHR. 2. The Characteristic of Transient Interference By transient interference is meant the partial structure echoes such as impulsive noise, lightning impulsive, meteor trail echoes and so on. It just partially pollutes the signal echoes, say, it only exists in several repeat periods and its existing time is very short, but its intensity is great. When the Doppler spectra is wide, it will cover the n target signal in frequency domain. The main parts of t c transient interference, meteor trail echoes and lightning impulsive, are introduced as follows. Meteor trail echoes is an electromagnetic wave which is reflected by meteor trail for OTHR. Meteors fall into two classes, meteor rain and paroxysmal meteor. Meteor rain only appears at some time in a year and the scene is pageantry, while paroxysmal meteor exists at every time. Therefore, the meteor trail echoes that OTHR receiving is generally generated by paroxysmal meteor. In every day, hundred millions of meteors enter the earth aerosphere. When these meteor enter the above floor aerosphere in a range from 80 to 120 km with the speed of 12 to 75 km/s, their kinetic energy transform to heat energy, and the atoms on meteors'surface are vaporized, at the same time ionizations happen in the atoms which have the same velocity as the meteor, so they leave over a trail of posi- tive ions and free electrons. lonization trail actually likes a thin and long gas column and meteor particles are at the head. The greater the quality of meteor entering is, the c tD greater the iron density in the ionization meteor trail is. As a result, it becomes the meteor trail which can reflect the electromagnetic wave. This trail lasts for a very short time, and in generally it just varies from a several tenths of a second to a few seconds. In some times, it could last for several minutes. The meteor trail can be further divided into over-density trail and lack-density trail. The over- density trail lasts for a longer time than the lack-density trail, but the over-density trail is very infrequent. The main characteristics of meteor trail echoes are as following : 1. The meteor trail echoes OTHR receives are mainly the reflects of lack-density trail, and they are generally in the range from 80 to 120 km. c 2. Because the meteor trail echoes last for a shorter time than the range from a few tenths of a second to one second, so it's Doppler spectra width varies from several tenths of a Hertz to one Hertz. At the same time, the movement of the meteor trail also leads to the movement of its Doppler spectra. 3. The meteor tail interference some times is very great, and it can not only enter from the main-lobe of the antenna beam but also from the side-lobe. And there are possibly not only direct echoes but also multipath echoes. ln radar beams, there are about 104-105 paroxysmal lack-density meteor trail echoes. Because of their high speed, they express a great intensity in the distance from 100 to 2 800 km, and the intensity is 10-20dB greater than that of the noise. Beside IOMHZ the intensity is the greatest while above IOMHZ it decreases greatiy. 4. Meteor trail likes a thin and long gas column small in diameter. Its diameter is generally about 0.5- 4.5m with length about 15-25 km (the length such as 30-40 km is only exist in 5% time, so the proba- bility is very small). Because the resolution ceil of the OTHR receiver is approximately 10 km, as a result meteor trail echoes generally exist in one or two range ceils and one or two beams, and if it enters from the side-lobe, it is possible to exist in many beams. 5. The lack-density meter trail echoes generally last for about a few hundred millisecond, while the coherent processing interval (CPI) of OTHR is generally approximately the order of ten second, so in generally, this kind of meter trail echoes only exist in a very short time comparing with CPI, and these are transient interference. 6. In some times, many trails appear by turns, in this condition the interference may last for one minute. The long existing meteor trail may be over-density trail echoes, and it is possible to be the mix of many trail echoes. 7. The chance to appear meteor rain is small, but once it should appear, the interference might be great. Lightning is the ionization and discharge that will break out when charges largely accumulate in the cloud and the voltage between cloud and ground (or cloud and cloud) increases to such a degree that it is larger than the strike air voltage. Lightning can be classified according to its happening position into two main classes : cloud lightning and ground lightning. Cloud lightning are those discharges that do not touch the ground, while ground lightning are those discharges that touch the ground. It is the ground lightning that has mainly influence on OTHR. Because a lightning lasts for 200-400 millisecond [5], compared with CPI which is the order of ten second, it is transient interference, and it is also plus interference. By the arisen frequency, the typical lightning impulsive is about one in a second or in five second. The difference of REE No' ? 3 Mars 2005 Transient Interference Excision in China Experimental OTHR lightning impulsive from meteor trail echoes is that it is not echo, so it has not correction with the transmitted signal, and the pulse compressing will not work on it. In addition to that, lightning impulsive is much longer than a general pulse period which is lOms-20ms. After pulse compressing, lightning impulsive exist in every range ceil and in a number of periods, but it comes in a direc- tion of space, so it mainly exists in one azimuth ceil. To the small-size airplane detecting, in the fulminous and lightning season, the lightning impulsive will make the use time of OTHR decreasing 25% [5]. 3. The Filter of Transient Interference The discussion above has mentioned that impulsive noise and meteor echoes are the main transient interfe- rence. Transient interference is greater than target echoes in intensity, but it is weaker than sea clutter or terrestrial clutter. Therefore in order to excise transient interference, it is deserved to be excised by gate-limit detecting under the condition that terrestrial clutter and sea clutter have been excised. There are different methods for terrestrial clutter environment and sea clutter environment. For ter- restrial clutter environment, FFT is firstly done on range and azimuth ceils of echoes to get the Doppler spectra of every ceil, next the spectra of terrestrial clutter is excised in frequency domain, thirdly the echoes return to time domain by IFFT, then the transient interference is detec- ted by a gate-limit detecting and excised from the origi- nal echoes which is not filtered terrestrial clutter, finally the excised clutter and target echoes are predicted by Burg linear prediction algorithm to smooth the data. While for sea clutter, its Doppler spectra is much wider than terrestrial clutter, but the close range and azimuth ceils have very strong coherence, while the coherence of transient interference is weaker. Sea clutter should be excised by the following steps : firstly the short time varying covariance matrix of received signal is estimated according to some close range and azimuth ceils, second- ly sea clutter is excised by MTI (which is got by the eigen-decomposition of covariance matrix and the pro- jection of signal on the maximums eigen-vector, which will be discussed behind) filter, thirdly the output is com- pared with a known gate-limit and transient interference is detected, fourthly the transient interference is excised from the original echoes which is not filtered terrestrial clutter, finally the excised clutter and target echoes are predicted by Burg linear prediction algorithm to smooth the data. According to this theory, it is possible to rene- wediy estimate the short time varying covariance matrix and renewedly filter, detect, excise and resume. In the following, MTI filter is firstly discussed. Suppose that there are P periods known clutter data with Q range ceils and M azimuth ceils, that is say that there are QM useable ceils. And suppose that a cell's former P periods clutter data at the nth time makes up of X,, = [x, (il - P + 1),,V (il - 1), X, (ii)]', where i is the nui-n- ber of the resolution range-azimuth ceil, and i = J,..., QM, so the time-varying (at nth time) covariance matrix estimated can be defined as: R = OM 1-1 i=1 Il T y Xi.nXi.H Xi.nXi.n (1) After eigen-decomposition of Rx, the eigenvalues are deserved to be arrayed from large to small, that is c AI > -2 >... > Ap'where those À whose value clearly large (suppose the number of them is/') are correspon- ding to the eigen-vectors that can expand the clutter sub-space (in generally, it is called signal sub-space, but under this condition, the signal is just the clutter). In this paper, the culler sub-space is expressed as SI, S2,... si-, While the other P - r eigen-vectors expand the noise sub-space, and GI, G,),.... Gp-r is used to express it. After the input vector X has been MTI filtered, that is say its part in the clutter sub-space is excised from X, the reflecting part of X reflecting on the clutter sub- ZD space can be computed as following : x. = , slxs (2) After MTI filter, the output is : xo =x-x.=X- Y sï xsj i-1 /=[ (3) It is showed that in the output the clutter has been excised, but almost all transient interference is held, and the next step is gate-limit detecting and excising tran- sient interference from the signal before filtering. In order to avoid the break at the position where clutter excised widening the clutter spectra, it is necessary to interpolate to the points where the clutter has been exci- sed according to the characteristic of clutter. Here Burg linear prediction algorithm is used to estimate the coef- ficient of AR model. Burg linear prediction algorithm is a algorithm to tD Zn c forecast the N th sampling xN with the former N sam- plings by estimatina the coefficient a,,,k-Of plings xo, xl,... xv c AR model, where k = 1,2,.... p. The following relation exits : x A- = a p.k _'-k (4) Where p is the order of AR model. The coefficient of AR model can be computed by Levison-Durbin formu- la, that is REE Nn 3 Mars 2005 M Dossqer LES RADARS al7.k = cll-l.k + CIP'Pap-1.1) -k Where a,,@,, isas following : : \--l (5) -21 (1' -1iib ''.v-i ii-j -' -1 2 + bl) _l li_l 2 1 l=p (6) When Eq. (6) is used, the denominator has the follo- wing relation : A'- ! DEN (IG) = E (f, /i) ? -L 2 + 1hpl,i-1) 2 n-p (7) 2 2 2 I) EN (P - 1) (1 - Cip-1.1) -l 2_) - + bp-l,.-P2 Where DEN (J » expresses the denominator of Eq. (6). By Eq. (5), Eq. (6) and Eq. (7), a group of coefficient fa,,,kj can be conformed, where the order of AR model is known before. The frame diagrams of whole processing to excise transient interference for different clutter circumstance are showed in Fig.l. In theory, the terrestrial clutter is narrow in spectra and the close ceils are weak in space coherence, so it is suit for directly filter in frequency domain. While the sea clutter is wide in its spectra and the close ceils are strong in space coherence, so it is suit for the covariance estimating to filter. In practical appli- cation, it is not entirely filtering the sea clutter and the terrestrial clutter but detecting the transient interferen- ce, which is the purpose. The processing frame diagram in Fie. 1 is suit to each other. c The meteor trail cchoes generally only exist in one range ceil and one beam, while the lightning impulsive exists in ail range ceils of one beam and the sea clutter or terrestrial clutter exists in all range ceils and ail azimuth ceils. The meteor trail echoes and lightning impulsive are great in one or all range ceils, but after averaging them in all range and azimuth ceils, their inten- sity is less than sea ctutter or terrestrial clutter, so Rx is mainly the sea clutter or terrestrial clutter covariance matrix. b FF'l'Clear terrestrial e J clutter dutter clutter comrios (D Lv-ï tecting transient Composc etecting interference 1 1 Excising the transient Interpolation Outputy (n) interference Clutter covariance Clutter sub-space c or matrix l ve matnx R vector S,.S,.....S Input.r(n) [Detecting transient L ,--,l interference Excising the transieit titerpolation output'i (ï) ..t f n erpo a Ion u pu1-'11 interference ( Frame dicigr (zii iiiide- the teri-esti-ial (littter ('it-ciiiiistciiice f Frarne diogram under the seo clutter circwnstonce Fig. 1. Fi-ciiiie diagi-aiii o.f detectiiig ajid excising transient interference. 4. The Processing of Real Data In the processing of real data from Chinese OTHR after the azimuth and range compressed with four azi- muth beams and cight range ceils, after the sea or terres- trial clutter is filtered, a meteor trail echo is found in a real data. Fig.2 (a) shows the echo of the eighth range ceil of the second beam, where the echo acutely changes at 6-8 second, and it can be concluded that there is a devia- tion from zero Hertz. Fig.2 (b) shows its spectra which has been divided by the length of the data. The partly amplification of Fig.2 (b) is showed in Fig.2 (g), and it is can be seen that at -0.4Hz and 0.5Hz there are two sea cluttcr's peaks with power 81.45dB and 83.5OdB. Between the peaks is there a peak with power 76.18dB at 0.1 Hz, and that is likely the terrestrial clutter spectra. The phase interference in ionosphere cali ses the move- ment of Doppler about 0. 1 Hz, at the same time the Burg peaks have been a litter widen, and the correction of the phase interference in ionosphere will be discussed in another paper. It is guessed that the azimuth and range ceil lies in the border of sea and ground, and the peak with power 73.54dB at 1.464Hz is likely the naval ship's spectra. Between -5.9Hz and -1.66Hz is there a jut that is found to be the spectra of meteor trail echoes in the behind processing. Fig.2 (c) shows the echoes after filte- ring clutter with eigen-vector algorithm, where between 0.625 second to 0.749 second is there an impulsive inter- ference which lasts for 124 millisecond. The figure also shows that the noise average power is about 65dB, while the max power of meteor trail echoes is 90dB which is about 15-25dB larger than the noise power. The compa- rison of Fig.2 (a) with Fig.2 (c) shows that the power of the sea or terrestrial clutter is about 20dB larger than that of noise. Fig.2 (d) shows the signal after excising the meteor trail echo with interpolation. Comparing Fig.2 (d) with Fig.2 (a), we can find that the acute changes at the position of the meteor trail echoes have been excised and the signal is continuous. In addition to those, in the pro- cessing it is found that in the same range ceil of the first beam also exist strong meteor trail echoes whose power is not stronger than that of the second beam, so does the same range ceil of the third beam. c REE J Mars 2005 Transient Interference Excision in China Experimental OTHR Figure 2.0 (a,b,c,d,efg,h). Excisioii the iiieteor trail echoes iiiteiferencef-oii t-ecil sigtial. 100 C-t S !' 10 v e R i FC O L .... , : 70 ---4 Z., 60 -1 Ti 1 1 60 401 1 1 1 1 1 0 2 4 6 8 1 () 0 2 4 6 8 10 time (s) mv 90 60 ----- ----- ----- ---- ----- - .... u g7o 1 Z, 6ü -M 1 50- i 0 1 1 1 1 1 o 2 4 6 9 1 () time (s) (a) Echo of the eighth raiige ceil of the secotid beaiii. (d) Signal after excising the meteor trail echo aftei-filteriiig the sea (terresti-ial) cltitter i,ith interpolatioii. 100 80 ro . ", r) o -- A 1 u fO 'n Fa s 40m'iL o -50 -30 -10 10 30 50 .10 10 Doppler (Hz) 100 80 0 :) la '-'60 - -1 .... 0 ; : t o 40 ) Mmm 20 01 1 1 1 1 1 -50 -30 -10 10 30 50 -10 10 Doppler (Hz) Spectra of the origiiial sigizal. (e) Spectra after excising the meteor trail echo. 10 () 90 O .. 0 fJ 60 50 1 p 40 1 1 11 0 2 4 6 8 10 time (s) luv go -.1 : : c " d 3 0 ' " o r ltll.i dkil I.â ÊIAÂ 40 01 -F-r--1 20 1m 1 -1 50 -10 10 -1 -10 10150 -10 10 Doppler (Hz) (c) The paroxyst-nal iiieteor trail echo. (f)Spectra after excisiiig the iiieteor trail echo vithout interpolation with interpolation. REE ? 3 Mars2005 LES RADARS 100 (D 0 CL vv 80 fI'l\ 1 J 1\ 60 f ,,.' ? 40 U M f 1J` i i <'.\,1f1 -- - i 'k f , ; il fi f. 4 J -'\f\, :.,J J \1\il h ff 20 0 -6 -4 -2 0 2 4 6 -2 0 2 Doppler (Hz) 100 x m w 0 CL BO 50 60, 3 t 40 Ii ri 1 20 0 f, O 6 -4 -2 0 Doppler(Hz) Ig) Partlr (ii-tl, (b) (li) P (ii-tl,- aiîîl ? lfï (,citioii (,g). All of these show that the three echoes are just the e same meteor trail echoes, but for the first and the third s beams, it is likely enter from the side-lobe. What is sho- wed in Fig.2 (e) is the spectra after excising the meteor trail echo without interpolation. In the figure is there a sect of zero, which leads to the spectra near the zero Hertz much widening, so it is unavailable for detecting the naval ships. Fig.2 (f) shows the spectra after excising the meteor trail echo with interpolation, and Fig2. (f) is the partly amplification of Fig.2 (h) near the zero Hertz. It can be found that the sea or terrestrial clutter'spectra is not widened. By comparing Fig.2 (f) with Fig.2 (b), it is found that the meteor trail echoes'spectra have been excised and they have a spectra width and a frequency movement, where the spectra width is 4.24Hz, the move- ment of the frequency center is 3.78Hz. In generally, the frequency of the naval ships is low and the echoes are weak in intensity. As a result if its frequency is between -5.9Hz aiid - 1.66, its spectra will be covered by that of the meteor trail echoes so that it can not be detected. In the original data of Fig.2, there is not naval ship target's spectra between -5.9Hz and - 1.66Hz. In ordei- to prove the efficiency of this algorithm, let us add a naval ship target with Doppler frequency -3.32Hz and power 60dB in the eighth range ceil of the second beam. Because its power is less about 5dB than the noise and the interference circumstance, the signal and spectra after adding a naval ship target are almost same as Fig.2 (a) and Fig.2 (b), where the naval ship's spectra is entirely submerged in that of the meteor. Fig.3 (a) shows the spectra after excising the meteor trail echo without interpolation, while Fig.3 (b) shows the spectra after excising the meteor trail echo with interpolation. After comparing the two figures, it will be found that in c ZD Fig.3 (a) it is difficult to detect the naval ship's spectra but in Fig.3 (b) its spectra is so clear that it can be detec- c ted with easy. 100 . -1 » 80 - 1 C2 60 - 1 1 a) 1 1 > 3' 0040 i 20. 1 o -50 -30 -10 10 30 50 -10 10 Doppler (Hz) f Sl ? ec-ti-ci Éifiei- e, ('isiiiç tlie iiieteoi- t-ciil e (-hoey H'ith iiiiei-I) olcitioii. 100 1 Il 80 Ship target C,5 60 - Q) - 0 " 5 f1. 40 20 -50 -30 -10 10 30 50 -10 10 Doppler(Hz) (b) Spectra after excisiig the iiieteoi- t-ail echoes vcithout interpolation. Fig.3. Excisioii tiie iiieteoi- trtiil eclioes iiiteii-eiice after (iclcliiig the iia,til shilq.igizal REE N 1 1 Miirs 2005 Transient Interference Excision in China Experimental OTHR 5. Conclusion A method for excising transient interference in OTHR has been discussed in this paper. In this method, the sea clutter or terrestrial clutter is firstly filtered out by eigen-dccomposition, then the position of the tran- sient interference is detected and the transient interfe- rence is excised from the original echoes, finally, the excised clutter and target echoes are predicted by Burg linear prediction algorithm. For this method there will be very small intluence on clutter spectra at the same time the transient interference is excised, which is advantaged to detect low-speed targets. In the result of the proces- sing of Chinese OTHR real data, the meteor trail echoes are found and the filter effect is testified. But the data Qot <-' is so few that the lightning impulsive is not exist. The fil- ter of lightning impulsive deserves further proof. c tD 6. Acknowledgment The authors t,,oiild like to thaiik the i-eiyiekt, j- theii- hell- ? Jitl coiiiiiieiits aiid siiggestioiis. The nuthor gmtefully nckr2owledges the support of K.C Wong education frrr.datiorz, Hor2 Kong. References [1] VV. Carrara, R.S. Goodman, R.M. Majewsk, "Spotlight Syntheüc Aperture Radar Signal Processing Algorithm ", Artech House, Boston London, 1995. [2] YI. Abramovich, S.J. Anderson, and I.S.D. Solomon, "Adaptive ionospheric distortion correction technique for HF sky wave radar " in Proceedings of the IEEE National Radar Conference, pp.267-272, IEEE Press, Piscataway, N.J.. 1996. [31 J. Parent, A. Bourdillon, " A Method to Correct HF Skywave Backscattered Signal of lonospheric Frequency Modulation ", IEEE Trans. Antennas Propag., Ap-35, 1988, pp.467-469. [4] S. J. Anderson, Y. 1. Abramovich, "A unified approach to detection, classification, and correction of/onosphere s- tortion ln HF sky vvave radar systems ", Radio Science, Volume 33, Number 4, pp.1 055-1067, July-August 1998. [5] J.R. Barnum,.E Simpson, " Over-the- Horizon Radar sensi- tivity Enhancement by Impulsive Noise excision ", 1997 National Radar Conference, pp.252-256. Glossary OTHR : Over-the-horizon radar HF: H ! gh-frequency PGA: Phase Gradient Algorithm PPP: Prominent point processing CPI : Coherent Processing Interval Information Le National Key Laboratory est spécialisé en traitement de signa avec des travaux sur des radars VHF, en STAP et reconnais sance de signature. REE Nn 1 Mars 2005