Research paper


Published on

Signal Processing Technique.

Published in: Education
  • Be the first to comment

  • Be the first to like this

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide

Research paper

  1. 1. Signal processing technique in Removing SignatureDistortion of Ultra-Wideband Radar and in Networkaided positioning. Ronak A Vyas B.TECH Electronics and Telecommunication SVKM’S NMIMS MUKESH PATEL SCHOOL OF TECHNOLOGY MANAGEMENT AND ENGINEERING, VILLE PARLE, MUMBAI ronak_vyasa_007@yahoo.in1. Introduction 32. Signal-Processing Technique to Compensate for Forward Motion 4 2.1 The SIRE Technique 4 2.2 Perfect Reconstruction of the Radar Signal With Stationary Radar Platform. 5 2.3 Phase and Amplitude Distortion in Reconstructed Radar Signal With Moving platform 6 2.4 Technique to Remove the Phase and Shape Distortions From Reconstructed 73. Network Aided positioning with signal processing 3.1 Positioning with Wireless LAN 9 3.2 Client-Based system design 10 3.3 Results 104. Summary 105. References 111
  2. 2. Signal processing technique to Remove SignatureDistortion of Ultra-Wideband Radar and in Network aidedpositioning.Abstract-- Ultrasonic detection and characterization of targets concealed by scatteringnoise is remarkably challenging. So a neural network (NN) coupled to split-spectrum processing (SSP) is examined for target echo visibility enhancement usingexperimental measurements with input signal-to-noise ratio around 0 dB The SSP-NNtarget detection system is trainable and consequently is capable of improving the target-to-clutter ratio by an average of 40 db. This system is exceptionally robust andoutperforms the conventional techniques such as minimum, median, average, geometricmean, and polarity threshold detectors. For real-time imaging applications, a field-programmable gate array (FPGA)-based hardware platform is designed for system-on chip(SOC) realization of the SSP-NN target detection system. This platform would be ahardware/software co-design system which will use parallel and pipelined multiplicationsand additions for high speed operation and high computational throughput. TheProcessing technique is used in removing signature distortion in Synchronous ImpulseReconstruction (SIRE) Ultra-Wideband Radar also. It would be helpful in increasingmobility, survivability, and lethality. However the targets won’t be stationary and therewould be phase and shape distortions. So this is countered with improvement in signal-to-noise ratio or focus quality. Hence this improvement results in Synthetic aperture radar(SAR) imagery. It can be applied to any time-based impulse radar system that experiencesthe relative motion between the radar and the targets during the data acquisition cycle.Ultrasonic signal processing methods can be used for detection of defects in many othercomposite materials. Ultra wideband processing technique could be effectively used inwireless world also. It is used to trace exact location of the caller, bomber, missilelauncher, etc. The location estimation can be done by scene analysis of RF or ultrawideband signal strength characteristics, which more or less works like pattern matchingin cellular location systems this could be a client based system design. Wirelesspositioning is becoming increasingly important.Keywords: Ultrasound inspection, Signal Processing, Synthetic aperture radar (SAR), forward imagesradar, Client based network aided system.1. IntroductionSignal Processing is enabling technology for the generation, transformation, and interpretation ofinformation. It comprises of applications related to processing information contained in many differentformats broadly designated as signals. Signal refers to any abstract, symbolic, or physical manifestation of2
  3. 3. information with examples that include: multimedia, sensor, communication, radar, biological, chemical,molecular, genomic, medical, data, or sequences of symbols, attributes, or numerical quantities.Signal processing uses techniques and algorithms for generating, transforming, transmitting, and learningfrom analog or digital signals. Signal processing involves techniques that improve our understanding ofinformation contained in received ultrasonic data. This paper describes a signal-processing technique thatremoves the phase and shape distortions from the radar signal, which are attributable to the motion of theplatform. This technique results in SAR imagery with significant improvement in focus quality and signal-to-noise level. This technique could be applied for any time-based impulse radar system that experiencesthe relative motion between the radar and the targets during the data acquisition cycle.2. Signal processing technique to compensate the forward motionARL (Army Research Laboratory) had developed a sampling technique that allows in-expensive ADCs todigitize wide bandwidth signals. This technique is called Synchronous Impulse Reconstruction (SIRE).2.1 The SIRE TechniqueThe ARL SIRE radar system employs an Analog Devices 12-bit 80-MHz ADC to digitized returned radarsignals. However, the ADC is clocked at the system clock rate of 40MHz. From the basic sampling theory,it is not possible to reconstruct the wide-bandwidth signal (300MHz to 3000 MHz) since this ADC rate ismuch slower than the required minimum Nyquist1 sampling rate (6000 MHz). However, by using thesynchronous time equivalence sampling technique, we can achieve a much higher equivalent sampling rate.The ADC sampling period is t; the value of this parameter in figure 1 is 25 nsec, which corresponds to ananalog-to-digital (A/D) sampling rate of 40 MHz. The number of samples for each range profile is denotedby N, which is equal to 7 in our example. This corresponds to a range swath of 30 m. The system pulserepetition frequency (PRF) is 1 MHz. The system pulse repetition interval i.e., the inverse of PRF, is 1micro-second ((mu)s).Each aliased (slowly sampled at A/D speed) radar record is measured M times (1024 in this example) andthe records are integrated to achieve a higher signal-to-noise level. After M repeated measurements of thesame range profile are summed, the first range (fast-time) bin is increased.This effective sampling rate is sufficient for the wide-bandwidth radar signal (300 MHz to3000MHz). After K groups of M pulses are transmitted and the return signals are digitized andsummed by the Xilinx Spartan1 field programmable gate-array (FPGA), this results in a radar recordof N.K samples with an equivalent of fast sample spacing of. The total time to complete one dataacquisition cycle is N.K.PRI. Please note that during the entire data acquisition cycle period, therelative position between the radar and the targets is assumed to be stationary. Table 1 summarizesthe parameters used by the SIRE data acquisition technique.. Figure 1 provides a graphical representation of the SIRE data acquisition technique.3
  4. 4. N=7 range gates @ A/D clock to cover the range swath Range Profile n=1 n=2 n=3 n=4 n=7 1 2 3 t: A/D Period M m=1 m=2 m= Averaging Factor:M= 1024 k= 1 2 3 4 193 Interleave Factor: K Reconstructed Waveform Figure 1. The ARL synchronous impulse reconstruction data acquisition scheme. (This is a modified and enhanced version of the equivalent time-sampling technique.)Table 1. Summary of radar parameters. Radar pulse repetition frequency (PRF) 1 MHz Radar pulse repetition interval (PRI) 1e-6 sec ADC sampling rate 40 MHz ADC sampling period 25 nsec Number of ADC (slow) range gates (N) 7 Interleaving factor (K) 193 Number of repeated measurements for averaging (M) 1024 Total number of range gates (N.K) 1351 Effective sampling period (time-equivalent) 129.53e-12 sec Effective sampling rate (time-equivalent) 7.72 GHz Total data acquisition time 197.6 msec------------------------------------------------------------------------------------------------------------------------1 Xilinx Spartan is a registered trademark of Xilinx, Inc.2.2 Perfect Reconstruction of the Radar Signal With Stationary Radar Platform4
  5. 5. Let us consider the first simulation case. The radar is situated 10 m away from a point target. Theradar transmits impulse signals to the point target. The receiver performs the data acquisition onreturned signals using the sampling technique described in previous section. After M.K pulses aretransmitted and received, the data acquisition cycle is completed and the radar waveform isreconstructed. In this case, since the radar is stationary, the reconstructed waveform is perfect.Figure 2 shows both the time and the frequency domain of reconstructed waveform with thestationary radar. Time domain of Reconstructed Wafeform Frequency domain of Reconstructed Waveform (Simulation) (Simulation) Radar is stationary during data Radar is stationary during data acquisition cycle acquisition cycle Distance from radar to target (meter) Frequency in Mhz Figure 2. Perfect reconstruction with stationary radar.2.3 Phase and Amplitude Distortions in Reconstructed Radar Signals With Moving PlatformWhen the radar is moving during the data acquisition cycle. This is the same simulation case as in previoussection except that the radar is moving toward the target at the speed of 5 miles per hour. In figure 3, weshow the reconstructed waveform of the moving case versus the stationary case. In the time domain plot offigure 3, we can notice a significant phase shift in the reconstructed waveform compared with the ideal case(stationary). The phase information is crucial for the SAR image formation process since radar signals arecoherently processed by the image former. In addition to the phase shift, the frequency domain plot of themoving case (figure 3) indicates that there is a distortion in the shape of the reconstructed waveform. Theamount of phase shift and shape distortions changes as the distance from the radar to the target varies.5
  6. 6. Figure 3 comparison of stationary and moving target.2.4 Technique to Remove the Phase and Shape Distortions From ReconstructedWaveformsThe phase and shape distortions in the reconstructed waveform can be explained as follow.Figure 3 shows the scheme when the radar is moving. For the reconstructed waveform, the effectivesampling period is Δm, which no longer has the same value as Δ in figure 1. After transmitting andreceiving the first M pulses, the radar starts the second group of M pulses with the timing offset of Δ fromthe previous group. During the data acquisition time for the first M pulses, in this case, the radar hastraveled a distance of d = v.M.PRI (2)In which v is the speed of the radar during this acquisition group of M pulses, and PRI is the pulse repetitionrate mentioned in section 2.1.Thus, the effective sampling period for the reconstructed waveform is Δm = Δ + t (3)From equation 3, the effective sampling period for the reconstructed waveform is varied with the radar’sinstantaneous speed. This generates the phase and shape distortion in the reconstructed waveform.From equations 2 to 3, we assume that the radar speed v is constant during the entire data acquisition cycle.This is not a bad assumption since the radar is moving slowly and its speed should not change.Let sm be the reconstructed waveform with phase and shape distortions attributable to the radar platform’smotion during the data acquisition cycle. Given the average speed of the radar platform at the time themeasurement is made, we want to compute the ideal reconstruction s from the distorted sm as if the radar isstationary during the data acquisition cycle. The system employs a differential global positioning system(GPS) system to measure the radar locations along the path. In addition, the GPS time stamps are recordedwith the radar data stream. With the radar coordinates and time stamps information from two successivelocations, the average speed of the radar at every location can be computed.6
  7. 7. Figure 4 Original signal and reconstructed signalFigure 5. The technique is applied to the simulated data (in figure 4). (This figure shows the perfectReconstruction with the processing technique.)Hence with signal processing technique the shape distortion is removed and its phase is aligned with theoriginal stationary signature.7
  8. 8. Figure 6 comparison of signal before and after processing3. Network aided positioning with signal processing techniquesThis system requires precise location of object which the radar is tracing and therefore there is need of somereliable and robust technology so wireless systems are taken into considerations. Network aided positioningdesign has different network topologies, physical layer characteristics, and media access control (MAC)layer characteristics require remarkably different positioning system solutions.3.1 Positioning with wireless LANWe are more or less interested in context-aware computing and location-aware services, which has led thedevelopment of wireless LAN-based indoor positioning systems, such as Bluetooth and Wi-Fi. WLAN-based positioning solutions mostly depend on signal strength utilization.8
  9. 9. 3.2 CLIENT-BASED SYSTEM DESIGNMany signal processing techniques have been proposed for location estimation for wireless networks.Location estimation is usually performed by scene analysis of RF or ultra wideband (UWB) signal strengthcharacteristics, which works much like pattern matching in location systems. Because signal strengthmeasurement is part of the normal operating mode of wireless equipment, as in Wi-Fi systems, no otherhardware infrastructure is required. A basic design utilizes two phases. First, in the offline phase, the systemis calibrated and a model is constructed based on received signal strengths at a finite number of locationswithin a targeted area. Second, during online operation in the target area, mobile units report the signalstrengths received from each access point (AP) and the system determines the best match between onlineobservations and the offline model. The best matching point is then reported as the estimated position.3.3 ResultsMany signal processing-based location algorithms include two stages: parameter measuring and positionestimation. For example, TOA (Time of arrival) can be determined either by measuring the phase of thereceived narrowband carrier signal or by directly measuring the arrival time of a wideband pulse. Radar andalso for sonar and GPS applications TOA estimation techniques have been widely used.Because the indoor multipath environment is very different from an outdoor environment, traditional TOAestimation algorithms, like the ML TOA (Maximum likelihood time of arrival) estimation technique, havebeen derived for applications where the radio propagation Channel can be simply modeled.Summary:This design of SIRE UWB SAR radar uses signal processing techniques for increased mobility,survivability, and lethality. The radar is based on time-domain wideband impulses. For this radar, ARLdesigned and implemented a data acquisition technique called SIRE that allowed us to employ relativelyslow ADC (40 MHz) to digitize wideband signals (>3000 MHz). However, the scheme assumed that theradar and targets are stationary during the data acquisition cycle when in reality; the target signatures didsuffer the distortions in phase and shape because of the radar motion. The phase error would lead tosignificant loss in target radar cross-sectional values in resulting SAR imagery. The shape errors woulddestroy the frequency contents of the targets and thus the ability to discriminate targets from other confusedclasses. This report described a signal-processing method that we designed to recover the accuracy of thetarget signatures that were affected by the radar motion. Signal Processing method has been applied tosimulated data and measured data from SIRE radar. With the measured radar data, There is a significantimprovement in the resulting SAR image. The radar cross section of targets improved from 5 dB to 13 dB.The correct shapes of target signatures are preserved. And various network aided technologies are beingused for precise location of object which the radar is tracing. Various networks can be used for this purposeand effective positioning is achieved without producing any distortions with help of Signal ProcessingTechniques.9
  10. 10. References:1. Nguyen, Lam; Wong, David; Ressler, Marc;Koenig, Francois; Stanton, Brian; Smith, Gregory; Sichina, Jeffrey; Kappra, Karl.Obstacle Avoidance and Concealed Target Detection Using the Army Research Lab Ultra-Wideband Synchronous ImpulseReconstruction (UWB SIRE) Forward Imaging Radar. Proceedings of SPIE, Detection andRemediation Technologies for Mines and Minelike Targets XII, Vol. 6553, April 2007.2. Ressler, Marc; Nguyen, Lam; Koenig, Francois; Wong, David; Smith, Gregory. The ArmyResearch Laboratory (ARL) Synchronous Impulse Reconstruction (SIRE) Forward-Looking Radar. Proceedings of SPIE,Unmanned Systems Technology IX, Vol. 6561, May 2007.3. Ressler, Marc; Nguyen, Lam; Koenig, Francois; Smith, Gregory. Synchronous Impulse Reconstruction (SIRE) Radar Sensorfor Autonomous Navigation. Army Science Conference, November 2006.4. Real-Time Versus Equivalent-Time Sampling, Tektronix, RAMAC/GPR Borehole radar, Mala Geoscience USA, Inc., www.malags.com6. Theory and Application of Digital Signal Processing by Lawrence R. Rabiner, Bernard Gold, Prentice Hall, Inc1975.7. J. Hightower and G. Borriello, “Location systems for ubiquitous computing,”IEEE Computer, vol. 34, no. 8, pp. 57–66, Aug. 2001.8. Special Issue on Wireless Geo-Location System and Services, IEEE Commun. Mag., vol. 36, no. 4, Apr. 1998.9. G.M. Djuknic, and R.E. Richton, “Geo-location and assisted GPS,” IEEEComputer, vol. 34, no. 2, pp. 123–125, Feb. 2001.10. [61] S. Gezici, Z. Tian, G. Giannakis, H. Kobayashi, A. Molisch, H.V. Poor, and Z. Sahinoglu, “Localization via ultra-wideband radios,” IEEE Signal Processing Mag., vol. 22, no. 4, pp. 70–84, July 2005.11. A. Sayed, A. Tarighat, and N. Khajehnouri, “Network-based wireless location,” IEEE Signal Processing Mag., vol. 22, no. 4,pp. 24–40, July 2005.12. N. Patwari, A.O. Hero, III, M. Perkins, N.S. Correal, and R.J. O’Dea, “Relative location estimation in wireless sensornetworks,” IEEE Trans. Signal Processing, (Special Issue on Signal Processing in Networks), pp. 2137–2148, Nov. 2002.13. P. Krishnan, A.S. Krishnakumar, W.H. Ju, C. Mallows, and S. Ganu, “A system for LEASE: System for location estimationassisted by stationary emitters for indoor RF wireless networks,” in Proc. IEEE Infocom 2004, Hong Kong, Mar. 2004, pp. 1001–1011.14. H.C. So and E.M.K. Shiu,“Performance of TOA-AOA hybrid mobile location,” IEICE Trans. Fund. Elect., Commun.Computer Sciences, vol. E86-A, no. 8, pp. 2136–2138, Aug. 2003.15. S. HAYKIN, ‘Neural Networks and Learning Machines’, Prentice Hall, 3rd Ed. New Jersey, 2008.16.Ressler, Marc; Nguyen, Lam; Koenig, Francois; Wong, David; Smith, Gregory. The Army Research Laboratory (ARL)Synchronous Impulse Reconstruction (SIRE) Forward-Looking Radar. Proceedings of SPIE, Unmanned Systems Technology IX,Vol. 6561, May 2007.17. Ressler, Marc; Nguyen, Lam; Koenig, Francois; Smith, Gregory. Synchronous Impulse Reconstruction (SIRE) Radar Sensorfor Autonomous Navigation. Army Science Conference, November 2006.18. Real-Time Versus Equivalent-Time Sampling, Tektronix, RTvET.html&FrameSet=oscilloscopes19 . RAMAC/GPR Borehole radar, Mala Geoscience USA, Inc., www.malags.com10
  11. 11. ------------------------------------------------------------------------------------------------Ronak A VyasB TechElectronics and TelecommunicationsThird yearMukesh Patel School Of Technology Management And EngineeringSVKM’s NMIMS University.Mumbai, IndiaEmail id: ; ronakvyasa007@gmail.comPhone No: +91- 9766798773.11
  12. 12. 12