00972099

594 IEEE JOURNAL OF OCEANIC ENGINEERING, VOL. 26, NO. 4, OCTOBER 2001

Underwater Acoustic Receiver EmployingDirect-Sequence Spread Spectrum and Spatial

Diversity Combining for Shallow-Water MultiaccessNetworking

Charalampos C. Tsimenidis, Oliver R. Hinton, Alan E. Adams, and Bayan S. Sharif

Abstract—This paper proposes an underwater adaptive-arrayreceiver structure that utilizes direct-sequence code division mul-tiple access and spatial diversity combining in order to achieve re-liable low-data rate multiuser communication in an asynchronousshallow-water network. The performance of the proposed receiverarchitecture has been verified by means of offline processing ofdata acquired during sea trials in the summer of 1999 in the NorthSea. Results show that this computationally efficient structure isnear–far resistant and provides successful multiuser operation inthe shallow-water channel.

Index Terms—Direct-sequence code-division multiple access,spread spectrum, shallow-water networking.

I. INTRODUCTION

D IRECT-SEQUENCE code-division multiple access (DS-CDMA) is a spread-spectrum (SS) technique that is often

utilized to achieve multiaccess communication. To be catego-rized as SS, a communications system must employ a trans-mission bandwidth that is considerably greater than the infor-mation rate. Utilization of bandwidth in this manner introducesa multiplicity of benefits, such as immunity against multipathand multiaccess interference suppression capability. However,the main advantage of DS-CDMA, particularly for underwateracoustic networks, is the support of asynchronous communi-cations [1]–[4]. Recently, the flourishing development and useof DS-CDMA in cellular-radio mobile communication systemsmotivated the application of this particular communication tech-nique in shallow-water acoustic networks.

The shallow-water acoustic channel is an exceptionallydifficult transmission medium that challenges the commu-nications methods available today. The principal difficultiesarise from multipath interference due to low-attenuated bottomand surface reflections associated with small grazing angles.These cause both long time-delay spread and large multipathamplitudes to be present in the received signal. In such ascenario, a system that employs DS-CDMA benefits from theimmunity that results from the utilization of the spreading codes

Manuscript received March 13, 2000; revised June 4, 2001. This work wassupported through the MAST-III Programme-D-G-XII-European Union withinthe LOTUS project under Contract MAS3-CT97-0099.

The authors are with the Underwater Research Group, Department of Elec-trical and Electronic Engineering, University of Newcastle Upon Tyne, New-castle Upon Tyne NE1 7RU, U.K.

Publisher Item Identifier S 0364-9059(01)09796-5.

Fig. 1. Transmitted packet structure.

or signatures used to distinguish between users. In addition,the large time–bandwidth product of the SS-classified signalwaveforms provides the system with the capability to recoverdiscrete multipath signal components that can be combined toallow diversity reception.

Furthermore, the use of the DS-CDMA spreading codes al-lows transmission between users to occur within the same fre-quency band and time slot. This presents an important featureof DS-CDMA permitting users to randomly access the channel.On the other hand, the blindness a user encounters during trans-mission can only be compensated by employing time-divisionmultiple access (TDMA) protocols. However, there are appli-cations such as the monitoring of instruments deployed in seawhere only half-duplex transmission is considered. The designof such a DS-CDMA communications system also needs to takeinto account the limited bandwidth available since the utilizationof SS signatures reduces considerably the effective transmissionrate. Thus, a designer is challenged to determine a balance be-tween system capacity and transmission efficiency.

Preceding research in the field of multiuser communicationsfor the shallow-water channel has demonstrated the feasibilityof DS-CDMA over signal-to-noise ratio (SNR) time-spreadchannels [5], [6]. In [7], both the centralized and decentral-ized versions of a single-element receiver were studied. Thedecentralized implementation of the receiver is based on thestructure analyzed in [8] and refers to a receiver algorithm thatdemodulates a single user by utilizing the information of theuser’s training sequence. In contrast, the centralized receiverrequires knowledge of all the users’ training sequences. In [9],the receiver structure initially examined in [7] was adapted tosupport spatial diversity processing.

In this paper, we confine ourselves to linear modulationschemes such as DS-CDMA quadrature phase-shift keying(QPSK), and important aspects of an adaptive array single-userreceiver. The paper is organized as follows. In Section II, the

0364–9059/01$10.00 © 2001 IEEE

TSIMENDIDIS et al.: UNDERWATER ACOUSTIC RECEIVER EMPLOYING DS SPREAD SPECTRUM AND SPATIAL DIVERSITY 595

Fig. 2. DS-QPSK signal constellation and transmitter architecture.

basic model of the communication system under considerationis outlined. Section III is devoted to the receiver architecture.Section IV presents experimental signal processing results.Conclusions and suggestions for further work are drawn inSection V.

II. COMMUNICATION SYSTEM DESCRIPTION

A two-user asynchronous shallow-water network is consid-ered. The direct-sequence SS communications system under ex-amination employs the signaling packet illustrated in Fig. 1. Thebinary phase-shift keying (BPSK)-modulated pseudonoise (PN)sequence is utilized to provide both a measure of the time delayinvolved in the transmission and to identify the presence of auser in the received signal. The latter is feasible due to the factthat a distinct PN sequence is assigned to each user. The BPSKpreamble is followed by an identical dual-channel QPSK-mod-ulated PN sequence spread by the user’s signature.

This signal overhead is required for the training of the adap-tive receiver structure needed to despread the received signal. Itshould be pointed out that the quadrature (Q) channel employsthe same but lag-displaced form of the PN training sequenceused by the in-phase (I) channel. The same principle holds forspreading waveforms too. The training sequences employed arespecifically selected to provide both sufficient autocorrelationprocessing gain, and low-valued crosscorrelation properties. Forthis purpose, commonly used preferred sequences were selectedexhibiting a three-valued cross-correlation function [10]–[12].The transmitter architecture that has been used to generate theDS-QPSK signal is depicted in Fig. 2.

III. RECEIVER STRUCTURE

The receiver is based on the adaptive correlator structure out-lined in [13]–[16] but modified to allow spatial-diversity com-bining processing [17], [18]. Analytically, the function of the re-ceiver algorithm can be subdivided into three processing stages.

A. Stage I: Signal Acquisition

The receiver front end is illustrated in Fig. 3. The receivedsignal is first bandpass-filtered in order to remove low-fre-quency signal disturbances. After I–Q mixing and chip-matchedfiltering, the acquired signal is sampled twice per chip. It isworth emphasizing that the preprocessing involved in this stage

Fig. 3. Front-end for theith receive array element.

Fig. 4. Correlation detector.

of the algorithm is identical for all the receive elements of thearray.

B. Stage II: Time-Delay Estimation and User Detection

The second stage operates on the acquired complex-base-band signal prior to any data-demodulation processing in orderto detect the presence of a user and to estimate signal param-eters required in the following stages. Among the most sig-nificant parameters is the initial time delay and Doppler fre-quency involved in the transmission between receiver and trans-mitter. However, since we adopt a noncoherent approach to thetime-delay estimation, compensation for the initial Doppler inthe received signal at this stage is not required.

The detection stage employed to provide the test statistic isdemonstrated in Fig. 4. The structure is generally known as acorrelation receiver. The received signal is processed through aPN-matched filter followed by a square-law envelope detector.The output of the envelope detector is compared to a prede-fined threshold, which is chosen to provide the desired false-alarm probability. In practice, the threshold is experimentallyset depending on the network scenario. Once the threshold isexceeded, the algorithm stores the following samples that spana window in time that approximates the delay spread of thechannel. The estimate of the initial time delay is the value thatcorresponds to the maximum magnitude in the stored samples.The correlator structure is known to be optimal for a single user


Fig. 5. Adaptive-array combiner.

in the presence of additive Gaussian noise but suffers signifi-cantly in performance in the presence of multiaccess interfer-ence (MAI) due to the near–far problem. Moreover, if notice-able Doppler shifts are present in the received signal, then thecorrelator structure performance is further degraded in terms ofcorrelation peak losses due to the narrow ambiguity function ofthe BPSK PN waveform. For the experimental result presentedin this paper, a waveform of 511-chips was selected which pro-vides a user with an advantage of 27 dB of processing gain.Given this length of code and the lack of significant movementbetween transmitter and receiver, the utilization of the correlatorstructure in conjunction with a PN waveform is justifiable. It is,however, important to realize that time delay must be estimatedindependently for each element of the receive array as deployed,due mainly to its dimensions and its specific geometric struc-ture.

C. Stage III: Demodulation

Once the presence of a user in the received signal is detected,the receiver algorithm proceeds over to the third stage in order todemodulate the transmitted information. The generic structure isdepicted in Fig. 5. Each receive element is assigned an adaptiveweight vector that can be implemented in the form of a finite im-pulse response (FIR) filter, where the complex taps are spaced athalf the chip rate. The adaptive algorithm simultaneously com-bines three functions. It performs adaptive equalization in orderto mitigate multipath effects present in the received signal. Ad-ditionally, it adaptively operates as a correlator/despreader; thus,the signal presented at its output is the recovered despread infor-mation. Finally, in the case of a multiuser scenario, it attemptsto mitigate MAI. To further analyze the operation of the adap-tive filter, it is constructive to consider its block diagram shownin Fig. 6.

At the output ofStage II, the complex-valued samples areshifted into a register to form an observation vector. Thelength of the observation vector is chosen to be twice thelength of the DS-CDMA signatures used by the communicationsystem. At each iteration, the contents of the shift register arecleared and a complete new set of samples is used to refill

Fig. 6. Adaptive correlator structure for theith branch.

the buffer. Given this, there is no need for explicit synchro-nization; the only requirements are the initial time delay andknowledge of the symbol duration so that the contents of theFIR filter can be properly updated between iterations of theemployed adaptive algorithm. The timing between transmitterand receiver is matched so that the right-most sample of thetaped-delay line corresponds to the first chip of the currenttransmitted data symbol. This ensures that the algorithm willkeep training on the correct data symbol even if the delayspread of the channel is greater than the length of the spreadingcode. Another prominent highlight of the suggested structure isthat exact knowledge of the spreading signature utilized by thetransmitter is not necessary. In contrast, the receiver preciselyrequires information about a predefined data training sequencein order to minimize the mean-squared error (MSE) duringoperation in training mode. When the achieved output SNRis sufficiently high, typically 5 dB to 10 dB, then the receiverswitches to decision-directed operation mode.

As previously indicated the main optimization criterionadopted is the minimization of the MSE. Mathematicallyexpressed,

(1)

Equation (1) indicates that in designing the adaptive FIRfilters, the goal is to find the matrix at time thatminimizes the quadratic cost function . Themost common iterative methods for minimization are thosebased on the recursive least square (RLS) and the least meansquare (LMS) algorithms. In the proposed receiver algorithm,the equalizer output is sampled at symbol rate while the input isclocked at the chip rate. This fact makes almost impractical theuse of algorithms, such as the RLS, that take advantage of thecyclic correlation between the equalizer contents at successiveoutput sampling times. We therefore concentrate on the LMSstochastic gradient algorithm. In our case, the tap weights ofthe adaptive equalizers are updated once per symbol accordingto the normalized LMS (NLMS) algorithm [19]. This variationof the standard LMS [20] is preferred due to the fact that theadaptation step can be additionally optimized according tothe inverse of the input signal power. The adaptive step sizerepresents a feature that is more attractive in practice since bothmultipath propagation and the presence of multiuser interfer-ence can dramatically affect the input-signal variations. The


Fig. 7. Shallow-water experiment site showing distinct transmit/receive positions.

error signal , which is constructed after spatial diversitycombing is given as

(2)

where is the number of receive elements and is thephase-corrected MMSE filter output of theth branch

(3)

and denotes the training sequence of the desired user.During the training period is precisely known, while in

the decision-directed mode it is derived from the diversity-combined output.

Strictly speaking, carrier-phase synchronization could beachieved implicitly by the complex equalizers. However, inthe presence of multiuser interference it is often preferableto use explicit, more rapidly converging, phase synchronizersembodied in the form of first-order decision-directed digitalphase-locked loops (DPLL). The phase estimate is obtained bydifferentiating the maximum log-likelihood function [21] withrespect to the phase

(4)


Fig. 8. Receive array structure.

TABLE IMAIN SYSTEM PARAMETERS

It should be emphasized that for the phase estimate of the first-order DPLL the observation interval (summation) must be re-stricted to the currently processed symbol only. The first-orderDPLL is considered to be optimal since the phase is expected tobe very slowly varying over subsequent transmitted symbols.

IV. EXPERIMENTAL PERFORMANCERESULTS

A. Experiment Description and Communication SystemParameters

To test the performance of the proposed receiver structure inmultipath and multiaccess interference scenarios, recorded datawere processed offline. The recorded signals were acquiredduring sea trials conducted in the summer of 1999, in theNorth Sea a few miles off the U.K. coast (Western Europe).Fig. 7 depicts the shallow-water experiment site. The distinctreceive/transmit positions give an impression of the simulatedmultiuser shallow-water network. The network was intended

Fig. 9. One-kilometer shallow-water channel impulse response evolution over6 s. Date: 05.07.99; time: 1325; sea state: 0; depth: 40 m; latitude: 5510.73;longitude: 01 25.83.

to cover ranges between 1–10 km whereby different scenarioswere set up to test both the near–far problem and the angularuser separating capability of receiver algorithms.

The receiver array employed was positioned in approximately40–50-m-deep water and is illustrated in Fig. 8. It consisted ofeight omnidirectional sensors comprising a horizontal plane of


Fig. 10. Velocity, salinity, and temperature profile (from left to the right).

four elements (sensors 3–6) and a vertical plane to provide bothbeamforming and spatial diversity reception. The two identicaltransmitters were positioned at 5 m distance from the seabed.The transmitting power was fixed at 190 dB rePa at a carrierfrequency of 9.6 kHz maximally band-limited at 4 kHz. Themodulation scheme employed was DS-CDMA QPSK. Table Isummarizes the major system parameters.

B. Channel Sounding and Environmental Data Monitoring

Prior to data transmission, a number of chirps were trans-mitted to determine typical impulse response properties of thechannel by correlation processing, and those for a 1-km channelare presented in Fig. 9. The detected shallow-water channel re-sponses exhibit delay-spread times of up to 15 ms for ranges be-tween 1 and 3 km. The figure shows the variation of the channelresponse over 6 s and it can be seen there is a very stable multi-path structure for these particular experiments. However, even inthese circumstances, each path is subject to independent phasefluctuations that must be tracked by the adaptive receiver.

Fig. 10 illustrates typical sound velocity, salinity, and tem-perature profiles extracted from measurements made during thesea trials. In the figure, the leftmost line is velocity, with salinityin the center and temperature at the right-hand side. The pres-sure scale is calibrated as depth in meters. The plots evidentlysuggest an approximately isothermal temperature profile. Thesound velocity profile changes at the topmost few meters ofwater are apparently due to variations in the salinity. The ray-tracing in Fig. 11 reveals a downward refracting channel and in-dicates that the major multipath propagation are not influencedby the time-varying nature of the sea surface.

C. Signal Processing

Offline signal processing and analysis of data acquired duringthe sea experiment was performed in order to determine the re-liability of the DS-CDMA receiver algorithm outlined in Sec-tion III. In the simulated two-user network scenario, User 1 waspositioned at a 3-km distance from the receiver while User 2was at 2 km. The angular separation between the users was45 degrees. It should be noted that for the demonstrated re-sults 15-chip spreading codes were employed for both users.Although it is possible to use longer spreading codes to increase

Fig. 11. Raytracing diagram.

Fig. 12. Received signal at sensor 1 (bottom).

the processing gain for a given source level and ambient noise,however, this does not necessarily improve performance dueto the convergence and tracking constraints imposed by rapidchannel variations.

Fig. 12 illustrates the received signal as acquired by thebottom sensor of the receive array. Although there is no overlap


Fig. 13. Channel impulse response for User 1.


in time in the transmission of the two users, this networkscenario is useful in order to evaluate the performance of thereceiver algorithm without any MAI. Figs. 13 and 14 depict thechannel impulse responses corresponding to the transmission ofUser 1 and User 2, respectively. The channel impulse responseswere estimated by employing the detection algorithm outlinedin Section III. In this network scenario, one can observe thedistinct multipath nature of the two channels. Evidently, User2 transmits through a multipath channel with an approximatedelay spread of 5 ms. In contrast the channel corresponding tothe transmission of User 1 exhibits little multipath propagation.The amount of expected inter-symbol interference (ISI) willbe most noticeable in the case of User 2. However, since inboth cases the duration of the CDMA code is greater than thedelay spread of the channel the effects of ISI will be negligible.Figs. 15 and 16 demonstrate demodulation results for both User1 and User 2, respectively. Error-free transmission is achievedwhereas the output SINR measured at the demodulator outputis 15.89 dB and 11.98 dB for User 1 and User 2, respectively.The SINR was calculated at the output of the spatial diversitycombiner according to [21]

(5)

(a) (b)

Fig. 15. Demodulation result for User 1: BER= 0/5000, SINR= 15.89 dB.(a) Mean-squared error. (b) OutputI–Q constellation.

(a) (b)

Fig. 16. Demodulation results for User 2: BER= 0/5000, SINR= 11.98 dB.(a) Mean-squared error. (b) OutputI–Q constellation

Fig. 17. Received signal at sensor 1 (bottom).

where is the number of the transmitted symbols within apacket excluding the symbols corresponding to the training se-quence. The receiver algorithm was switched to the decision-di-rected mode after 200 symbols.

Fig. 17 depicts the signal received by sensor 1 in the nextscenario under consideration. Here, there is overlap in time inthe transmission of the two users. Specifically, the signal trans-mitted by User 2 arrives in the middle of the transmission ofUser 1, well after the end of the training period. This representsa worst-case situation for User 1 due to the fact that User 1 op-timizes the receiver parameters without any multiple-access in-terference, thus when the signal of User 2 arrives the receiver al-gorithm can rely only on the rejection capability of the CDMAcode. In contrast, User 2 optimizes its receiver parameters by




(a) (b)

Fig. 20. Demodulation results for User 2: BER= 0/5000, SINR= 14.73 dB.(a) Mean-squared error. (b) Output I–Q constellation.

utilizing the training sequence in the presence of User 1. Hence,it is expected to perform nearly unaffected in the presence ofUser 1. The advantage of User 2 in terms of transmit power overUser 1 is estimated to be 2.26 dB. Figs. 18 and 19 illustrate thechannel impulse responses for User 1 and User 2, respectively.The demodulator output is shown in Figs. 20 and 21. A closerlook at the resulting SINR levels reveals that the resulting degra-dation due to MAI remains for both users in the range of 1 dB.

(a) (b)

Fig. 21. Demodulation results for User 2: BER= 0/5000, SINR= 10.87 dB.(a) Mean-squared error. (b) Output I–Q constellation.

Fig. 22. Output SINR versus element combination for the weak user.

To further analyze the performance of the proposed receiver,let us consider Fig. 22. Demonstrated in this figure is theachieved SINR at the output of the diversity combiner as afunction of the number of the receive elements employedin the demodulation algorithm. Results shown in this figurefor the weak user represent an average over 10 min of trans-mission. This corresponds to approximately 10transmittedsymbols. Here, the proposed algorithm is compared withthe performance of a simple matched filter receiver utilizingequal-gain combining in conjunction with a first-order DPLLand automatic gain control. Evidently, the performance ofthe suggested adaptive-correlator receiver outperforms theconventional matched filter in all cases. A small degradation inperformance is observed in the single-element situation wherethe convergence of the adaptive algorithm can be critical; buteven in this case the performance is very similar. Furthermore,in order to investigate the contribution of the diversity arrayin the performance of the receiver algorithm the length ofthe adaptive correlator is shortened. Demonstrated is the casewhere only the first seven chips of the spreading code areused corresponding to approximately half the symbol energy.Clearly, the performance in terms of the achieved output SINR


drops significantly for the single-element receiver. However,the sensitivity to the length of the correlator becomes lesssignificant as the number of the receive elements employedincreases.

V. CONCLUSION

An adaptive-array receiver architecture that utilizesDS-CDMA and spatial diversity combining has been proposedfor reliable low data rate multiuser communications in an asyn-chronous shallow-water network. The most outstanding featureof the receiver algorithm is the approach that integrates threefundamental communications functions into one structure:despreading, equalization, and multiaccess interference rejec-tion. Moreover, the only information required by the receiveris the knowledge of the distinct training sequences utilized todetect the presence of a user. These are required to train theadaptive equalizers at the beginning of the transmission. Theperformance of the receiver was evaluated by means of offlinesignal processing of experimental data. The demonstrated re-sults were compared against the performance of a conventionalmatched-filter receiver employing equal-gain combining. Inall cases the proposed receiver algorithm outperformed theconventional one in both single and multichannel arrangements.

Further investigation is continuing to both enhance the relia-bility of the existing algorithms and to evaluate the performanceof the receiver structure by employing different spreading codelengths. In addition, higher modulation schemes are being ex-amined in conjunction with longer spreading codes in order tooffset the reduction in data rate that is dramatically degraded bythe utilization of DS-CDMA. Finally, a real-time implementa-tion of the proposed receiver algorithm in a three-node shallow-water network was successfully implemented in the summer of2000 where the performance of the proposed receiver was testedagainst TDMA protocols and multiuser/multistage approachesto the multiaccess problem.

REFERENCES

[1] R. L. Pickholtz, D. L. Schilling, and L. B. Milstein, “Theory ofspread-spectrum communications—A tutorial,”IEEE Trans. Commun.,vol. COM-32, pp. 855–884, May 1982.

[2] M. K. Simon, J. K. Omura, R. A. Scholtz, and B. K. Levitt,Spread Spec-trum Communications Handbook. New York: McGraw-Hill, 1994.

[3] S. Glisic and B. S. Vucetic,Spread-Spectrum CDMA Systems for Wire-less Communication. Boston, MA: Artech House, 1997.

[4] G. Woodward and B. S. Vucetic, “Adaptive detection for DS-CDMA,”Proc. IEEE, vol. 88, pp. 1413–1434, July 1998.

[5] C. Boulanger, G. Loubet, and J. Lequepeys, “Spreading sequences forunderwater multiple access communications,” inProc. Oceans ’98,Nice, France.

[6] G. Loubet, V. Capellano, and R. Filipiak, “Underwater spread-spectrumcommunications,” inProc. Oceans ’97, Halifax, Canada.

[7] Z. Zvonar, D. Brady, and J. Catipovic, “Adaptive detection for shallow-water acoustic telemetry with cochannel interference,”IEEE J. OceanicEng., vol. 21, pp. 528–536, Oct. 1996.

[8] M. Stojanovic, J. Catipovic, and J. G. Proakis, “Phase coherent digitalcommunications for underwater acoustic channels,”IEEE J. OceanicEng., vol. 19, pp. 100–111, Jan. 1994.

[9] M. Stojanovic and Z. Zvonar, “Multichannel processing of broadbandmultiuser communication signals in shallow water acoustic channels,”IEEE J. Oceanic Eng., vol. 21, pp. 156–166, Apr. 1996.

[10] W. W. Peterson and E. J. Weldon, Jr.,Error-CorrectingCodes. Cambridge, MA: MIT Press, 1972.

[11] M. B. Pursley, “Numerical evaluation of correlation parameters for op-timal phases of binary shift-register sequences,”IEEE Trans. Commun.,vol. COM-27, pp. 1597–1604, Oct. 1979.

[12] D. V. Sarwate, “Crosscorrelation properties of PN and related se-quences,”Proc. IEEE, vol. 68, pp. 593–619, May 1994.

[13] U. Madhow and M. L. Honig, “MSSE interference suppression for di-rect-sequence spread-spectrum CDMA,”IEEE Trans. Commun., vol.42, pp. 3178–3188, Dec. 1994.

[14] S. L. Miller, “An adaptive direct-sequence code-division multiple-accessreceiver for multiuser interference rejection,”IEEE Trans. Commun.,vol. 43, pp. 1746–1754, Feb. 1995.

[15] , “Training analysis of adaptive interference suppression fordirect-sequence code-division multiple-access systems,”IEEE Trans.Commun., vol. 44, pp. 448–495, Apr. 1996.

[16] C. N. Pateros and G. J. Saulnier, “An adaptive correlator receiverfor direct-sequence spread-spectrum communication,”IEEE Trans.Commun., vol. 44, pp. 1543–1552, Nov. 1996.

[17] C. C. Tsimenidis, O. R. Hinton, B. S. Sharif, and A. E. Adams, “Anadaptive array DS-CDMA receiver for a shallow-water asynchronousmultiuser network,” inProc. Oceans ’99, Seattle, WA, Sept. 1999.

[18] , “Spread-spectrum based adaptive array receiver algorithms for theshallow-water acoustic channel,” inProc. IEEE Oceans 2000, RhodeIsland, RI, Sept. 2000.

[19] S. Haykin,Adaptive Filter Theory. Englewood Cliffs, NJ: Prentice-Hall, 1996.

[20] B. Widrow, “The complex LMS algorithm,”Proc. IEEE, vol. 63, pp.719–720, April 1975.

[21] J. G. Proakis,Digital Communications. New York: McGraw-Hill,1995.

Charalampos C. Tsimenidiswas born in Greece in1971. He received the Dipl.-Ing. degree in electricalengineering from the University of Munich, Munich,Germany, in 1997 and the M.Sc. degree in commu-nications and signal processing from the Universityof Newcastle, Newcastle Upon Tyne, U.K., in 1998,and is currently working toward the Ph.D. degree inelectrical engineering at the same university.

He is currently employed as a Research Associateat the Underwater Acoustic Group at Newcastle Uni-versity. His research interests include multiuser de-

tection, estimation and spread-spectrum communications.

Oliver R. Hinton was born in 1947 in U.K. Hereceived the B.Sc.(Eng.) and the Ph.D. degrees inmicrowave circuits from University College London,London, U.K., in 1968 and 1972, respectively.

He is currently a Professor in Signal Processingand Head of the Department of Electrical and Elec-tronic Engineering at the University of Newcastle,Newcastle Upon Tyne, U.K. He has been a visitingscholar to Colorado State University, Fort Collins,and to Stanford University, Stanford, CA. Hisresearch interests are in signal processing, digital

communications, and subsea acoustics.He has published over 100 papers in academic journals and conferences,

and has managed over 27 research contracts of total value over £3M. He wasan invited lecturer on the CEC Advanced Course on Acoustical Oceanography,a U.K. Representative at the SERC/MTD N+N Meeting in Brighton in March1994 on U.K./U.S. discussions for a joint Research Programme for CleanerSeas, was an EC MAST III U.K. Expert Reviewer (invited) in November1996, and has been member of various Conference Committees includingOCEANS’98. Prof. Hinton is a member of the Institution of ElectricalEngineers (IEE) and has served on IEE Professional Group Committees E5and E15.


Alan E. Adams was born in Stoke-on-Trent, U.K.,in 1949. He received the B.Sc. (Hons.) degree inelectronic engineering from the Polytechnic ofNorth Staffordshire, Staffordshire, U.K., in 1970 andthe research M.Sc. degree from the University ofDurham, Durham, U.K., in 1977 for the developmentof a microprocessor based imaging system.

He is presently a Senior Lecturer with the Depart-ment of Electrical and Electronic Engineering of theUniversity of Newcastle, Newcastle Upon Tyne, U.K.He is the author of an undergraduate text on micro-

processor systems. His current research interests center on the use of acousticsignals in the marine environment; for communication, imaging and environ-mental measurements.

Mr. Adams is a corporate member of the Institution of Electrical Engineers.

Bayan S. Sharif received the Bachelor and Doc-torate degrees from Queen’s University, Belfast,Belfast, Ireland, and from Ulster University, Ulster,Ireland, in 1984 and 1988, respectively.

In 1989, he was a Research Fellows at Queen’sUniversity of Belfast, where he worked on parallelprogramming algorithms for two-dimensionalsignal-processing applications. He joined NewcastleUniversity, Newcastley Upon Tyne, U.K., in 1990as a lecturer in electronic engineering, where he iscurrently a Professor of Digital Communications

and head of the Communications and Signal Processing Research Group. Hisresearch interests are in DSP algorithms for digital communications and imageprocessing.

Prof. Sharif is a Chartered Engineer and a member of the Institution of Elec-trical Engineers.

00972099

Documents