DIRECT MAPPING OVSF-BASED TRANSMISSION SCHEME FOR UNDERWATER ACOUSTIC MULTIMEDIA COMMUNICATION

Journal of Marine Science and Technology, Vol. 18, No. 3, pp. 413-418 (2010) 413 DIRECT MAPPING OVSF-BASED TRANSMISSION SCHEME FOR UNDERWATER ACOUSTIC MULTIMEDIA COMMUNICATION Chin-Feng Lin*, Jiang-Yao Chen*, Ya-Ju Yu*, Jung-Ting Yan*, and Shung-Hyung Chang** Key words: OVSF, assignent, underwater acoustic ultiedia counication. ABSTRACT Underwater counication networks are currently being developed for use in underwater control systes, underwater teleetry systes, and underwater iage transission systes. The transission bandwidth and transission rates of an underwater counication syste are low, and its transission delay is high. In this paper, we propose a direct apping orthogonal variable spreading factor (OVSF)-based transission syste for underwater acoustic ultiedia counication. An iportant feature of this transport architecture is that a assignent echanis, an OVSF schee, a direct apping strategy, an schee, and are eployed for transitting essages that can tolerate various bit error rates. For confiring the effectiveness of this transport architecture, we perfor a siulation using easured data of a G.729 audio and a JPEG2000 iage. The siulation results show that the proposed syste can achieve axiu transission data rates or iniu transission. In addition, the OVSFbased transport architecture is a feasible platfor in underwater acoustic ultiedia counication. I. INTRODUCTION A ajor part of the earth s surface is covered by the ocean. Deep oceans are popularly known as inner space. Huans have often enthusiastically explored this inner space in order to gain a better understanding of the underwater world. A popular and interesting area of research is underwater acoustic counication [1-3, 5-6, 8-9, 11, 15-18, 20-21, 24]. At present, underwater acoustic transission syste can be used to Paper subitted 10/29/08; revised 06/29/09; accepted 06/30/09. Author for correspondence: Chin-Feng Lin (e-ail: lcf1024@ail.ntou.edu.tw). *Departent of Electrical Engineering, National Taiwan Ocean University, Keelung, Taiwan, R.O.C. **Departent of Microelectronic Engineering, National Kaohsiung Marine University, Kaohsiung, Taiwan, R.O.C. issue coand/control instructions in underwater systes, to easure transission data of underwater easureent instruents such as sonar systes, and to transit underwater iages. It is siilar to land counication using ultiple access schees. Exaples of ultiple access schees include frequency division ultiple access (FDMA), tie division ultiple access (TDMA), spread spectru technique, and orthogonal frequency division ultiplexing (OFDM). However, due to a harsh underwater environent, underwater counication faces soe unique probles such as liited bandwidth, high and variable propagation delays, high bit error rates (BER), and severe frequency-selective distortion caused by ultipath propagation. Spread spectru counication is considered as a candidate technique for use in future obile underwater acoustic networks. Its advantages include efficient use of bandwidth, flexible guard band as copared to that in FDMA, and less stringent synchronization requireents as copared to those in TDMA. In the past, various studies have been conducted to investigate the technologies used in underwater acoustic counication systes. Authors proposed an underwater -array receiver structure in which direct-sequence code division ultiple access (DS-CDMA) and spatial diversity cobining are used for achieving reliable low-data rate ultiuser counication in an asynchronous shallow-water network [24]. In [5], the authors found that in underwater acoustic networks, the perforance of a direct-sequence spread spectru syste is better than frequency-hopping spread-spectru syste in ters of the transission BER. In [9], the authors use DS-CDMA and ulti-carrier CDMA to achieve reliable ultiuser counication in asynchronous shallow-water acoustic networks. In DS-CDMA, spread data are transitted at a single carrier frequency. In contrast, in MC-CDMA, a set of carrier frequencies is used to achieve frequency diversity. In [9], the authors found that for low -to-noise ratios (SNRs), the perforance of the DS-CDMA syste was better than that of the MC-CDMA syste and for SNR values higher than 20 db, the MC-CDMA syste outperfored the DS-CDMA syste due to the frequency diversity of the forer syste. In [21], the authors proposed a ultichannel detection ethod and an efficient channel-estiation-based ultiuser detection ethod for wideband underwater acoustic CDMA counication. In

414 Journal of Marine Science and Technology, Vol. 18, No. 3 (2010) Audio Iage Audio Iage user 1 user n G.729 JPEG2000 G.729 JPEG2000 OCPN MODEL OCPN MODEL data audio iage data audio iage a 1 (t) d 11 (t) OVSF1 Direct b 1 (t) Mapping d 1 (t) d 12 (t) a n (t) OVSFn d 1 (t) Direct b n (t) Mapping d n (t) d 2 (t) s 11 (t) s 12 (t) s 1 (t) s 2 (t) u 1 s 11 (t) u 1 s 12 (t) μ s 1 (t) μ s 2 (t) Underwater channel odel r 11 (t) r 12 (t) SNR_data SNR_audio SNR_data Fig. 1. Proposed OVSF-based underwater acoustic transission syste. [16], the authors developed a dynaic ultiple access protocol for different underwater acoustic network architectures and traffic scenarios; this protocol efficiently shares scarce underwater channel bandwidth by fully leveraging CDMA ediu access properties. They also ipleented a novel closed-loop distributed algorith to jointly set the optial values of transit and code length that cobat for the near-far effect. In previous studies [4, 10, 11-15], we exained wireless ulti-edia counication, obile teleedicine systes, and underwater acoustic ulti-edia counication. For exaple, in [15], we studied a single input single output (SISO) OFDM-based transission schee for underwater acoustic ultiedia counication. Further, in [11], we studied a transission schee based on SISO orthogonal variable spreading factor (OVSF) codes for use in underwater acoustic ultiedia counication. In this study, we extend this previously work [11], and develop an underwater acoustic ultiedia transission syste in which the OVSF, a assignent echanis, an schee, direct apping (DM) and an schee are eployed for transitting essages with low error probability and at low. We jointly set the transit, code length, type, and coding type for underwater ultiedia counication and to cobat for the near far effect. In addition, we show that the proposed underwater acoustic ultiedia counication syste can achieve axiu transission data rates or iniu transission. II. DIRECT MAPPING OVSF-BASED TRANSMISSION SCHEME FOR UNDERWATER ACOUSTIC MULTIMEDIA COMMUNICATION OVSF-based transission is one of the various ultiple access techniques eployed in an underwater acoustic counication syste. An advantage of such a syste is that it can facilitate robust data transission. The transission architecture of the proposed syste is shown in Fig. 1. Fro this figure, we can observe that the proposed syste can effectively deal with various types of s in an underwater environent. When users play ultiedia in the receiver of the underwater acoustic counication syste, various ultiedia objects are synchronized. The tie dependent relations aong various objects as well as their play throughput and transission data rates for various objects are estiated. A odel called the object-coposition Petri-net (OCPN) [25] can describe the teporal relationships aong the various aspects of ultiedia inforation, such as its type, size, throughput requireents, and duration for which it is presented in the receiver. Such aspects can then be delivered by the proposed underwater acoustic ultiedia counication syste. Usually, the quality of service (QoS) requireents for various essages in a ultiedia syste are different. Here, we assue that the acceptable BERs for audio, and iage are 10 3, and 10 4, respectively. A 1500 kbits test iage is copressed by a Joint Photographic Experts Group 2000 (JPEG2000) to produce a 100 kbits iage bit strea. Further, a 1330 kbits test audio is copressed by a G.729 to produce a 96 kbits audio bit strea. These iage, and audio bit streas are introduced in the OCPN odel. Two types of iage and audio are used in our OVSF-based underwater acoustic counication syste for transitting audio, and iage bit streas, respectively. The total nuber of audio and iage channels can easily be estiated by the OCPN odel. OVSF codes are applied in spread data to cobat channel fading and achieve the required QoS for underwater acoustic ultiedia transission. There is only one code in OVSF with spreading factor 1, and the first OVSF code is 1. The second OVSF code (1, -1) is generated fro the first OVSF code, and the spreading factor of the second code is 2. In addition, the ith OVSF

C.-F. Lin et al.: Direct Mapping OVSF-based Transission Schee for Underwater Acoustic Multiedia Counication 415 code C with an spreading factor of 2 i 1, and the ith OVSF code set having 2 i 1 OVSF codes can be generated as Ci 1Ci 1 if xi = 0 Ci = Ci 1( Ci 1) if xi = 1 In our proposed syste, we use the carrier sense ultiple access/collision avoidance (CSMA/CA) protocol [7], and the lengths of the OVSF code c(t) are 8, 16, 32, and 64 for various. CSMA/CA belongs to a class of protocols called ultiple access schees. In CSMA/CA, a station wanting to transit data ust first listen to the shared channel for a predeterined aount of tie so as to check for any activity on the channel. If the channel is sensed as idle, that the station is peritted to transit. If the channel is sensed as busy, the station has to defer its transission. The baseband transission d (t) obtained using the th OVSF code for the th user is expressed as (1) d () t = a () t b () t (2) Here, b (t) is the data, which consists of a sequence of rectangular pulses of duration T, for the th user, and b (t) is the th OVSF spreading sequence for the th user. The output sequence of d (t) is (d 1, d 2, d 3, d 4, ), and then, it is subjected to DM [26]. DM is a transission echanis in the IEEE 802.11n standard, and it assigns the sequence (d 1, d 2, d 3, d 4, ) sequence to N parallel units. Further, a serial-to-parallel converter also splits this sequence, d 1, d 2, d 3, d 4, into N branches. Thus, we describe d 1 (t) = d 1 (t), = 1, 3, 5,, and d 2 (t) = d 1 (t), = 2, 4, 6,. The first transducer is used to transit s 1 (t), the chip sequence of d 1 (t). Siilarly, the second transducer is used to transit s 2 (t), the chip sequence of d 2 (t). DM can achieve high transission data rates for underwater acoustic ultiedia counication. The received lth hydrophone for th user is expressed as N = n nl + n= 1 γ () t u PS ()* t h () t n () t (3) Here, constant P is the transission of the transitter; µ is the transission factor, and 0< µ 1. Further, h il is the channel ipulse response of the ith transducer and lth hydrophones; We assue perfect channel estiation, and n (t) is the additive white Gaussian noise (AWGN) for the th user, obtained by the receiver using the lth hydrophones. L is the total nuber of hydrophones in the receiver. The SNR obtained by receiver for the th user is given by SNR = L l = 1 L l = 1 Er 2 { ( t)} En 2 { ( t)} (4) Thus, the decision on the output is L D () t = r () t a() t (5) l = 1 bˆ () t = dec {D ()} t For D (t), the receiver akes a decision according to specified thresholds. We suarize our assignent algorith as follows: Step 1: On the basis of the output inforation of the OCPN odel, obtain throughputs of audio and iage essages for underwater transission. Step 2: Select an appropriate paraeter and ode to satisfy the requireents for an underwater syste. Step 3: Assign the transission factor µ, 0 < µ 1, for audio or iage. Step 4: Measure the received SNR values in the cases of audio or iage. Step 5: If the easured SNR of the received is larger than the threshold SNR that can yield the required BER, update the transission factor to µ = µ and go to Step 4. Otherwise, go to Step 6. Step 6: Increase the transission factor to µ = µ +. If µ > 1, reselect both the error protection paraeter and the ode, and go to Step 3. If µ 1, go to Step 4 and repeat the reaining steps. The value of paraeter depends on the variation in channel fading. The greater the variation in channel fading, the larger is the value of. In addition, the saller the variation, the larger is the aount of saved. The lengths of the OVSF codes for audio, and video LC_a and LC_v are 8, 16, 32, and 64. The channel coding rates are obtained using 1/2 (561, 753) and 1/3 (557, 663, 771) convolution codes. The possible length of OVSF codes, type, and channel coding for audio, video, and data are (64, 1/2, BPSK), (64, 1/2, QPSK), (32, 1/2, BPSK), (32, 1/2, QPSK), (16, 1/2, BPSK), (16, 1/2, QPSK), (8, 1/2, BPSK), (8, 1/2, QPSK), (64, 1/3, BPSK), (64, 1/3, QPSK), (32, 1/3, BPSK), (32, 1/3, QPSK), (16, 1/3, BPSK), (16, 1/3, QPSK), (8, 1/3, BPSK), and (8, 1/3, QPSK). The initial for audio, video, and data is 1/30, and the axiu is 1. III. SIMULATION We perfored a siulation to deonstrate the functionality of the proposed OVSF-based underwater transission syste. In this siulation, we used, DM, CSMA/ CA, OVSF codes, assignent algorith, and the schee. We used the underwater chan-

416 Journal of Marine Science and Technology, Vol. 18, No. 3 (2010) BER 10 0 10-1 10-2 10-3 10-4 10-5 Power Weighting 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1-20 -19-18 -17-16 -15-14 -13-12 -11-10 10-6 -26-24 -22-20 -18-16 -14-12 -10-8 -6 SNR (db) Fig. 2. Bit error rate perforance of proposed OVSF-based underwater acoustic transission syste with 1/2 (561, 753) coding and BPSK. (*: spreading factor (SF) = 8, O: spreading factor (SF) = 16, ½: spreading factor (SF) = 32, : spreading factor (SF) = 64) nel odel developed by Zhang et al. [26] in this siulation. In this odel, the transission range is 1000, the frequency of a carrier wave is 11.5 khz, and the bandwidth is 3.90625 khz. The transitter contains two transducers and the receiver contains two hydrophones. The BER perforance of the proposed underwater transission syste is shown in Fig. 2. We use BPSK, K = 9 1/2 (561, 753) and 1/3 (557, 663, 771) convolution codes with soft decoding [23], and a 2 2 DM strategy for the transission of audio and iage. The spreading factor for the OVSF codes are 8, 16, 32, and 64. The greater the length of spreading codes, the lower is the transission BER. The transission for the proposed syste in the case of BERs of 10 3, and 10 4 for audio and iage, respectively, under different noise conditions is shown in Fig. 3 as a function of the AWGN (N o ). Here, the length of OVSF code is 8. Fro Fig. 3, we can observe that the higher the noise, the higher is the transission. Further, less restrictions on the transission BER result in low transission. The transission of audio s is saller than that of iage s. Further, the transission of 1/2 convolution codes is larger than that of 1/3 convolution codes. We consider No = -10 db. The lengths of OVSF codes are 8, 16, 32, and 64. Saller length of OVSF codes result in low transission. Tables 1 and 2 show the descend and transission data rates for various lengths of OVSF codes, types, and channel coding types for audio and iage s, respectively. Here, No is -10 db. The axiu transission data rates is 976 bits for the bandwidth of 3.90625 khz, and the axiu descend is 80%. The descend is defined as Rb Rb ub 100% (6) R b No (db) Fig. 3. Transission perforance of proposed OVSFbased underwater acoustic transission syste with OVSF = 8. (, dotted line: 1/3 convolution, BPSK, audio; O, dotted line: 1/3 convolution, QPSK, audio; *, dotted line: 1/2 convolution, BPSK, audio; ½, dotted line: 1/2 convolution, QPSK, audio;, line: 1/3 convolution, BPSK, iage; O, line: 1/3 convolution, QPSK, iage; *, line: 1/2 convolution, BPSK, iage; ½, line: 1/2 convolution, QPSK, iage) Table 1. shows the descend and transission data rates for various the length of OVSF codes, types, and channel coding types for audio s. the length of OVSF code convolution code transission data rates (bits) descend (%) 8 BPSK 1/2 488 18/30 40 8 QPSK 1/2 976 25/30 16.67 8 BPSK 1/3 324 14/30 53.3 8 QPSK 1/3 648 18/30 40 16 BPSK 1/2 244 14/30 53.3 16 QPSK 1/2 488 18/30 40 16 BPSK 1/3 162 11/30 63.3 16 QPSK 1/3 324 14/30 52.72 32 BPSK 1/2 122 9/30 70 32 QPSK 1/2 244 13/30 56.67 32 BPSK 1/3 80 7/30 76.67 32 QPSK 1/3 162 11/30 63.3 64 BPSK 1/2 60 7/30 76.67 64 QPSK 1/2 120 9/30 70 64 BPSK 1/3 40 6/30 80 64 QPSK 1/3 80 7/30 76.67 Where R b and µ b are the transission data rate and the transission, respectively, for audio or iage s. Fro Tables 1 and 2, we observe that the proposed syste can achieve axiu transission data rates or iniu transission. Figure 5 shows the perforance of a G.729 audio in the proposed underwater transission syste with the assignent echanis. The ean square error (MSE) of the original and received

C.-F. Lin et al.: Direct Mapping OVSF-based Transission Schee for Underwater Acoustic Multiedia Counication 417 Table 2. shows the descend and transission data rates for various the length of OVSF codes, types, and channel coding types for iage s. the length of OVSF code convolution code transission data rates (bits) descend (%) 8 BPSK 1/2 488 19/30 36.67 8 QPSK 1/2 976 26/30 13.33 8 BPSK 1/3 324 17/30 43.3 8 QPSK 1/3 648 19/30 36.67 16 BPSK 1/2 244 15/30 50 16 QPSK 1/2 488 19/30 36.67 16 BPSK 1/3 162 12/30 60 16 QPSK 1/3 324 15/30 49.38 32 BPSK 1/2 122 10/30 66.67 32 QPSK 1/2 244 14/30 53.33 32 BPSK 1/3 80 8/30 73.33 32 QPSK 1/3 162 12/30 60 64 BPSK 1/2 60 8/30 73.33 64 QPSK 1/2 120 11/30 63.33 64 BPSK 1/3 40 7/30 76.67 64 QPSK 1/3 80 8/30 73.33 Aplitude 1 0.8 0.6 0.4 0.2 0-0.2-0.4-0.6-0.8 0 1 2 3 4 5 6 7 8 Tie (s) Fig. 5. Received and decoded G.729 audio s with assignent echanis. (MSE = 0.0033) 0.9 0.8 Power Weighting 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 10 20 30 40 50 60 70 The spreading factor of OVSF code Fig. 4. Transission perforance of proposed OVSFbased underwater acoustic transission syste with No = -10 db. (O, dotted line: 1/3 convolution, BPSK, audio;, dotted line: 1/3 convolution, QPSK, audio; *, dotted line: 1/2 convolution, BPSK, audio; ½, dotted line: 1/2 convolution, QPSK, audio; O, line: 1/3 convolution, BPSK, iage;, line: 1/3 convolution, QPSK, iage o; *, line: 1/2 convolution, BPSK, iage; ½, line: 1/2 convolution, QPSK, iage) audio s is 0.0033. As shown in the figure, the quality of the audio is good. Figure 6 shows the received JPEG2000 iage when the assignent algorith is used. The peak SNR (PSNR) of the received iage is 42.8 db. Figure 7 shows the received JPEG2000 iage in the case that the assignent algorith is not used. Fro Figs. 6 and 7, we observe that it is feasible to use the proposed Fig. 6. Received and decoded JPEG2000 iage s using assignent algorith. (PSNR = 42.8 db) assignent algorith is feasible in the underwater acoustic counication syste. Further, we can conclude that the proposed syste can efficiently transit audio and iage s. VI. CONCLUSION In this paper, we proposed an underwater acoustic ultiedia transission schee in which the CSMA/CA protocol, OVSF code, DM transission strategy, a assignent algorith,, and an schee were eployed. We also perfored a siulation to deonstrate the functionality of the proposed syste; the siulation results were in good agreeent with the theoretical discussion. The proposed syste can efficiently

418 Journal of Marine Science and Technology, Vol. 18, No. 3 (2010) Fig. 7. Received and decoded JPEG2000 iage s without assignent algorith. (PSNR = 24.1 db) transit audio and iage s. In addition, it can achieve axiu transission data rates or iniu transission. ACKNOWLEDGMENTS The authors acknowledge the support of the grant fro the National science Council of Taiwan NSC 93-2218-E-019-024 as well as 96-2221-E-002-016 and the valuable coents of the reviewers. REFERENCES 1. Baggeroer, A., A survey of acoustic teleetry, OCEANS, pp. 48-54 (1981). 2. Baggeroer, A., Acoustic teleetry An overview, IEEE Journal of Oceanic Engineering, pp. 229-235 (1984). 3. Calvo, E. and Stojanovic, M., Efficient channel-estiation-based ultiuser detection for underwater CDMA systes, IEEE Journal of Oceanic Engineering, pp. 502-512 (2008). 4. Chang, P. R. and Lin, C. F., Design of spread spectru ulti-code CDMA transport architecture for ultiedia services, IEEE Journal on Selected Areas in Counication, pp. 99-111 (2000). 5. Freitag, L., Stojanovic, M., Singh, S., and Johnson, M., Analysis of channel effects on direct-sequence and frequency-hopped spread-spectru acoustic counication, IEEE Journal of Oceanic Engineering, pp. 586-593 (2001). 6. Gui, J. H., Kong, J., Gerla, M., and Zhou, S., The challenges of building scalable obile underwater wireless sensor networks for aquatic applications, IEEE Network, pp. 12-18 (2006). 7. IEEE P802.11 Wireless LANs, TGn sync proposed technique specification, IEEE 802.11-04/0889r6 (2005). 8. Kilfoyle, D. B. and Baggeroer, A. B., The state of the art in underwater acoustic teleetry, IEEE Journal of Oceanic Engineering, pp. 4-26 (2000). 9. Konstantakos, D. P., Tsienidis, C. C., Adaas, A. E., and Sharif, B. S., Coparison of DS-CDMA and MC-CDMA techniques for dual-dispersive fading acoustic counication networks, IEE Proceeding Counication, pp.1031-1038 (2005). 10. Lin, C. F. and Chang, K. T., A assignent echanis in ka band OFDM-based ulti-satellites obile teleedicine, Journal of Medical and Biological Engineering, pp. 17-22 (2008). 11. Lin, C. F., Chang, S. H., Chen, J. Y., and Yan, J. T., A assignent echanis for underwater wireless ultiedia, Proceeding of MTS/IEEE Ocean (2008). 12. Lin, C. F., Chang, W. T., Lee, H. W., and Hung, S. I., Downlink control in ulti-code CDMA obile edicine syste, Medical & Biological Engineering & Coputing, pp. 437-444 (2006). 13. Lin, C. F., Chen, J. Y., Shiu, R. H., and Chang, S. H., A Ka band WCDMA-based LEO transport architecture in obile teleedicine, Teleedicine in the 21st Century, edited by Lucia Martinez and Carla Goez, Nova Science Publishers, pp. 187-201 (2008). 14. Lin, C. F. and Li, C. Y., A DS UWB transission syste for wireless teleedicine, WSEAS Transactions on Systes, pp. 578-588 (2008). 15. Lin, C. F., Shih, C. H., Chen, C. P., Leu, S. W., Wu, J. K., Tseng, C. H., Hung, H. S., Lu, F. S., Parinov, I. A., and Chang, S. H., An OFDM-based transission schee for underwater acoustic ultiedia, WSEAS Transactions on Counications, pp. 343-352 (2009). 16. Popili, D., Melodia, T., and Akyildiz, I. F., A CDMA-based ediu access control for underwater acoustic sensor networks, IEEE Transactions on Wireless Counications, pp. 1899-1909 (2009). 17. Proakis, J. G., Sozer, E. M., Rice, J. A., and Stojanovic, M., Shallow water acoustic networks, IEEE Counications Magazine, pp. 114-119 (2001). 18. Rutgers, D. P. and Akyildiz, I. F., Overview of networking protocols for underwater wireless counications, IEEE Counications Magazine, pp. 97-102 (2009). 19. Sozer, M., Stojanovic, M., and Proakis, J. G., Underwater acoustic networks, IEEE Journal of Oceanic Engineering, pp. 72-83 (2000). 20. Stojanovic, M., Recent advances in high-speed underwater acoustic counications, IEEE Journal of Oceanic Engineering, pp. 125-136 (1996). 21. Stojanovic, M. and Freitage, L., Multichannel detection for wideband underwater acoustic CDMA counication, IEEE Journal of Oceanic Engineering, pp. 685-695 (2006). 22. Technical Specification Group Radio Access Network Multiplexing and Channel Coding (FDD), 3rd Generation Partnership Project (3GPP), Release 7, pp. 25-212 (2007). 23. Technical Specification Group Radio Access Network Spreading and Modulation (FDD), 3rd Generation Partnership Project (3GPP), Release 7, pp. 25-213 (2007). 24. Tsienidis, C. C., Hinton, O. R., Adas, A. E., and Sharif, B. S., Underwater acoustic receiver eploying direct-sequence spread spectru and spatial diversity cobining for shallow-water ultiaccess networking, IEEE Journal of Oceanic Engineering, pp. 594-603 (2001). 25. Woo, M., Prabhu, N., and Ghafoor, A., Dynaic resource allocation for ultiedia services in obile counication environents, IEEE Journal on Selected Areas in Counication, pp. 913-922 (1995). 26. Zhang, J., Zheng, Y. R., and Xiao, C., Frequency-doain equalization for single carrier MIMO underwater acoustic counications, Proceeding of MTS/IEEE Ocean (2008).