TRANSMITTING CAPSULE ENDOSCOPE IMAGES WITH WIRELESS LAN AND SMART ANTENNA SYSTEMS

BIOMEDICAL ENGINEERING- TRANSMITTING CAPSULE ENDOSCOPE IMAGES WITH WIRELESS LAN AND SMART ANTENNA SYSTEMS SHAOU-GANG MIAOU 1, SHIANN-SHIUN JENG, CHEN-WAN TSUNG, CHIH-HONG HSIAO 1, TAH-YEONG LIN 3 1 Deparmen of Elecornic Engineering, Chung-Yuan Chrisian Universiy, Chung Li, Deparmen of Elecrical Engineering, Naional Dong Hwa Universiy, Hualien, 46 Biomed. Eng. Appl. Basis Commun. 006.18:46-54. Downloaded from www.worldscienific.com by 148.51.3.83 on 04/05/18. For personal use only. 3 Chung Shan Insiue of Science and Technology, Taiwan ABSTRACT Capsule endoscopy gradually replaces radiional endoscopy in some applicaions and becomes a sae-of-he-ar ool o deec he problems of inesines. When capsule endoscopy is used, a paien swallows a capsule-like micro-camera. An image sequence is hen aken by he capsule endoscope and ransmied o a receiver carried by he paien. Evenually, hese image daa will be ransmied o a deskop compuer and examined by a docor. To sar he diagnosis earlier and avoid limiing he paien's movemen, an on-line wireless ransmission for his las link is desirable. For his link, WLAN (Wireless Local Area Nework) sandard is a good candidae due o is high enough daa rae and commercial availabiliy. However, wireless links ofen resul in ransmission errors ha are unaccepable for medical relaed applicaions, including medical image ransmission. In his paper, we propose a WLAN sysem wih smar anenna o ransmi he capsule endoscope images and evaluae is performance. The simulaion resuls demonsrae ha uilizing he smar anenna can enhance he error resilien capabiliy of he WLAN over an error prone wireless channel and provides a much reliable daa link for he ransmission of capsule endoscope images han he original sandard. Biomed Eng Appl Basis Comm, 006(Ocober); 18: 46-54. Keywords: wireless local area nework; smar anenna; capsule endoscope image 1. INTRODUCTION In he radiional endoscopy examinaion, a paien usually feels uncomforable because he paien mus swallow a long laparoscope. Recenly researchers have developed a new echnology called capsule endoscope [1-3], where a camera and a radio ransmier as small as a capsule is swallowed by he paien. Then an Received: March 9, 006; Acceped: Augus, 006 Correspondence: Shaou-Gang Miaou, Professor Deparmen of Elecornic Engineering, Chung-Yuan Chrisian Universiy, Chung Li 303, Taiwan E-mail: miaou@wavele.el.cycu.edu.w image sequence will be aken by he capsule endoscope, followed by he radio ransmission of he resuling inesine images o a receiver carried by he paien. This process will coninue for abou 6 o 8 hours and he receiver will receive ens of housands images. Evenually, hese images need o be ransmied o a deskop compuer and examined by a docor o diagnose he problems of inesines. One way of ransmiing he image daa o he deskop compuer is off-line and wired. This is also he only commercially available soluion a presen. However, his mehod has following disadvanages. Firsly, i will consume much ime when ens of 33

Biomed. Eng. Appl. Basis Commun. 006.18:46-54. Downloaded from www.worldscienific.com by 148.51.3.83 on 04/05/18. For personal use only. 47 housands images are ransmied. For example, if he sysem produces wo 4-bi full-color images per second and repea ha procedure for 8 hours, he size of produced images daa is abou 11 GB if he image size is 56 56. When we ransmi hese images daa using USB.0, i will ake abou 3 minues. For he image size of 51 51 in he fuure, we need o spend 1 minues o ransmi he image daa. Secondly, he diagnosis process can only begin a leas 8 hours laer afer he paien swallows he capsule endoscope. This is inconvenien for boh he paien and he docor. For mos cases capsule endoscopy is used for small inesines. When a capsule endoscope eners he large inesines, i sill akes up o 1 o hours o leave he paien's body. However, he exac ime is difficul o predic because i can vary a lo from one paien o anoher. Thus, we need o wai enough bu uncerain ime before we can download he image daa from he receiver o he deskop compuer. If we can use an on-line ransmission mehod, we can save much ime and begin he diagnosis process much earlier. For on-line ransmission, real ime is a basic requiremen. When real-ime ransmission is used, wireless ransmission is a good choice o ransmi he image daa if we wan o avoid he limiing of paien's movemen. In a commercially available capsule endoscope, he daa is ransmied in frames per second and he frame size is 56 56. In his case, a leas 56 56 8 1 mega bis need o be ransmied per second for real-ime consideraion. If he size of he ransmied image is 51 51, he daa rae will be up o 4 Mbps. If he frame rae increases o 30, he daa rae becomes 15 Mbps. Wih he sae-of-he-ar echnology, Blueooh (version.0) and Infrared (version 1.1) can provide a daa rae higher han 1 Mbps, bu he effecive ransmission disance o achieve such a rae is very limied, say 1 meer, which is no very desirable. On he oher hand, he wireless local area nework (WLAN) can provide much higher daa raes in a furher disance. The IEEE 80.11 sandard series is one of he mos popular WLAN sandards [4]. I replaces he cable-based Eherne in many applicaions because i has he advanages of easy se-up and high mobiliy. IEEE published he IEEE 80.11g sandard in 003 [5-6]. The maximum daa rae of IEEE 80.11g is 54 Mbps, and he operaing radio frequency is in he Indusrial, Scienific, and Medical (ISM) band. When he IEEE 80.11g is considered for he ransmission of image daa, he characerisic of elecromagneic wave propagaion and he inerference coming from oher wireless devices may inroduce serious ransmission errors. For he medical applicaion, he error olerance for he received image is low because any error may influence he diagnosis made by he docor. Moreover, Vol. 18 No. 5 Ocober 006 he docor may lose his or her confidence in he sysem when oo many ransmission errors occur. The IEEE 80.11g sandard was no designed o ransmi medical relaed daa, including medical images. An enhanced version mus be provided for ha purpose. In his paper, we aemp o apply a smar anenna echnique o he IEEE 80.11g sysem o improve is performance and reliabiliy for medical image ransmission. The inended arge image is he capsule endoscope image. The smar anenna echnique can enhance he desired signal and suppress he inerference simulaneously by direcing he radio frequency (RF) radiaion beam o desired users and suppressing he anenna radiaion paern in he direcion of he inerference. In order o change he radio radiaion paern, muliple anennas wih digial signal processing algorihms are required. More deails on smar anenna echniques are given in Secion II, where an inroducion o IEEE 80.11g is given as well. Secion III shows he simulaion resuls of his work. Finally, a conclusion of his work is given.. PROPOSED SYSTEM.1 Sysem Overview Afer a paien swallows a capsule endoscope, he paien's inesine images are wirelessly ransmied o he sorage device carried by he paien. For example, in one capsule endoscope sysem [1], he paien wears special clohes ha have he necessary equipmen inside. The equipmen is basically a baery-powered micro-compuer wih wireless communicaion capabiliy. The proposed capsule endoscope communicaion sysem is shown in Fig. 1. The capsule endoscope ransmis he daa o he sorage device in a mobile paien uni. Then, he mobile paien uni uses he IEEE 80.11g sysem o ransmi he daa o he deskop compuer wih a buil-in smar anenna sysem. Thus, only he uplink (one-way communicaion) of he IEEE 80.11g sysem needs o be considered in his work. Fig.1. The block diagram of he proposed sysem. (Rx: receiver, Tx: ransmier) 34

BIOMEDICAL ENGINEERING- 48 Biomed. Eng. Appl. Basis Commun. 006.18:46-54. Downloaded from www.worldscienific.com by 148.51.3.83 on 04/05/18. For personal use only.. IEEE 80.11g IEEE 80.11g suppors several baseband modulaion schemes, including DSSS/CCK (Direc Sequence Spread Specrum/Complemenary Code Keying), OFDM (Orhogonal Frequency-Division Muliplexing), PBCC (Packe Binary Convoluional Code) and CDMA (Code Division Muliple Access) - OFDM (opional). In his work, we adop OFDM as he baseband modulaion scheme of IEEE 80.11g. I uilizes 64-poin FFT (Fas Fourier Transform) in he OFDM modulaion. Only 5 sub-channels are used for daa ransmission. In hese 5 sub-channels, 4 subchannels are reserved for he ransmission of pilo signals. These pilo signals are used for channel esimaion and Doppler shif esimaion and compensaion. There are 8 ransmission daa raes and differen raes usually correspond o differen ransmission modulaion schemes. A lis of IEEE 80.11g parameers is shown in Table I. When he ransmied daa is processed by he physical layer, some conrol symbols are padded o he ransmied daa, such as PLCP (Physical Layer Convergence Proocol) header and PLCP preamble. Afer adding he conrol symbols, he ransmied daa is called he PPDU (Physical Proocol Daa Uni) packe. The PPDU consiss of hree fields, including PLCP raining field, SIGNAL field and DATA filed. The field arrangemen is shown in Fig.. The PCLP raining symbol consiss of 10 shor raining symbols and long raining symbols. The SIGNAL field conains 4 bis, where he 4-bi rae subfield specifies one of he 8 ransmission raes; he 1-bi ransmission lengh subfield is used o describe he ransmission daa lengh and he range of his value is from 1 o Table I. A lis of IEEE 80.11g parameers PLCP Header Daa Reserved Lengh Pariy Tail Service PSDU Tail Pad 4 bis 1 bi 1 bis 1 bi 6 bis 16 bis 6 bis 1bi PLCP Preamble 1 symbols SIGNAL One OFDM Symbol PPDU DATA Variable Number of OFDM Symbols Fig.. The PPDU packe forma of IEEE 80.11g. 4095. In he DATA field, he service subfield conains 16 bis, where he firs o he 6-h bis are used for synchronizaion of he scrambler a he receiver, and he 7-h o he 15-h bis are reserved for fuure applicaions. The DATA field also includes PSDU (PLCP Service Daa Uni), Tail, and Pad Bis..3 Smar Anenna Sysems A smar anenna sysem mainly consiss of muliple anennas and digial signal processing (DSP) echniques. I can improve communicaion performance by exploiing he space diversiy. There are wo ypes of DSP algorihms for smar anennas - he swiched beamforning and he adapive beamforming. In he swiched beamforming approach, an appropriae beam-selecion algorihm is used o consruc a direcional paern wih muliple narrow beams. Swiched beamforming echniques are well developed and no sophisicaed beamforming mechanism is required in his approach. In adapive Parameer Value NSD: Number of daa subcarriers 48 NSP: Number of pilo subcarriers 4 NST: Number of subcarriers, oal 5 (N SD + N SP ) F: Subcarrier frequency spacing 0.315 MHz (0 MHz/64) TFFT: IFFT/FFT period 3. s (1/ F) TSIGNAL: Duraion of he SIGNAL BPSK-OFDM symbol 4 s (T GI + T FFT ) TGI: GI (Guard Inerval)duraion 0.8 s (T FFT /4) TGI: Training symbol GI duraion 1.6 s (T FFT /) TSYM: Symbol inerval 4 s (T GI +T FFT ) TSHORT: Shor raining sequence duraion 8 s (10T FFT /4) TLONG: Long raining sequence duraion 8 s (T GI + T FFT ) TPREAMBLE: PLCP preamble duraion 16 s (T SHORT + T LONG ) 35

Biomed. Eng. Appl. Basis Commun. 006.18:46-54. Downloaded from www.worldscienific.com by 148.51.3.83 on 04/05/18. For personal use only. 49 beamforming, he smar anenna sysem adaps iself o he dynamically changing RF requiremen so ha i can direc beams o desired users and suppress he anenna paern in he direcion of he inerference, where boh users and inerference sources are no fixed. The adapive beamforming approach generally needs more complicaed signal processing work and higher cos [7-1]. Fig. 3 shows he basic implemenaion archiecure for a smar anenna sysem. The sysem is composed of an anenna array and several ransceivers. The signal received/ransmied by he anenna array is convered o/from baseband signals hrough he RF/IF unis and processed wih smar uplink and downlink algorihms. These algorihms deermine he uplink weigh vecors for performing beamforming on he received signals as well as he downlink weigh vecors for he ransmied signals. In he uplink, he smar uplink algorihm acquires he spaial signaure (o be defined shorly) of a desired user and adops he beamforming echniques o combine he signals. In he downlink, he signals are processed by he smar downlink algorihm according o he spaial signaure acquired in he uplink, convered ino RF signals and ransmied o he desired user. In he endoscopy applicaion, we only need o consider he performance evaluaion of smar anennas for he uplink of IEEE 80.11g. Thus, he downlink will no be addressed furher in he following paragraphs. Fig. 3. Archiecure of a smar anenna sysem. Assume ha he ransmied signal a he mobile paien uni is carried in plane wave and he anenna array a he deskop compuer is composed of M- elemen omnidirecional anennas arranged in a uniform and linear fashion. The received signals conain boh direc pah and mulipah signals. The array response vecor o a ransmied signal s 1 () from a direcion-of-arrival (DOA) is given by 1 1 Vol. 18 No. 5 Ocober 006 a ( ) [1, a ( ), a ( ),..., a ( )] T M (1) where a i ( ) is a complex number denoing he ampliude gain and phase shif of he signal a he (i+1)-h anenna relaive o ha a he firs anenna and T is he ranspose operaor. For a uniform and linear anenna array wih separaion D in free space, he array response vecor due o he signals can be wrien as D D a ( ) 1,exp j f sin,...,exp j f sinm 1 c c () where f and c denoe he carrier frequency and speed of ligh, respecively [13]. In a ypical wireless scenario, he anenna array is comprised of azimuhally broad coverage, even when omnidirecional elemens are used. Therefore, i no only receives a signal s 1 () propagaed along he direc pah bu also many mulipah echoes from differen DOA's. Wih his in mind, he oal signal vecor received by he anenna array can be wrien as x() a( ) s () a( ) s () n() 1 1 where L is he oal number of pah (conaining direc pah and mulipah); he complex number k describes he phase and ampliude difference beween he kh mulipah and he direc pah; n() is he background noise; and 1 1 l l 1 l a s () n() is referred o as he spaial signaure (SS) associaed wih Source one, s 1 (). From Equaion (3), we know ha he spaial signaure a 1 is a linear combinaion of direc pah and mulipah signals and depends on he direc and mulipah DOA's. We can hus uilize a direcionfinding algorihm, such as DFT, ESPRIT [14], and MUSIC [15], o find he direc pah and mulipah DOA's. According o hese DOA's, we can calculae he complex weigh o be used in he beamforming algorihm [16-18] over he received signals o obain he desired signal and suppress he inerference signal. The DOA based beamforming algorihms uilized in his work are described as follows: (1) Dominan DOA (DDOA) approach: The approach firs capures he uplink spaial signaure and hen finds he DOA's of he received signals using subspace based echniques such as MUSIC and ESPRIT. The ampliudes L L a a ( l ) 1 1 l1 l (3) associaed wih T 36

Biomed. Eng. Appl. Basis Commun. 006.18:46-54. Downloaded from www.worldscienific.com by 148.51.3.83 on 04/05/18. For personal use only. BIOMEDICAL ENGINEERING- he DOA's are also esimaed. The DOA wih he maximum ampliude, l, is seleced and is array response vecor a( 1 ) is chosen as he uplink weigh vecor. () Pseudoinverse DOA (PIDOA) approach: The approach is similar o he dominan DOA echnique excep ha we ake he pseudoinverse of he array response vecors of all he DOA's excep for he DOA of he desired user. This mehod places nulls in all DOA's excep for our desired user in order o minimize he inerference. To illusrae his mehod, suppose ha he kh mobile uni has one direc pah signal and a k k mulipah signal wih DOA's 1 and, respecively. k k Thus, is uplink spaial signaure is a k a 1. ka. To simplify our presenaion, we assume ha here are only wo independen sources, s 1 () and s (). If he weigh vecors for hese wo signals are w 1 and w, hen he received signal afer beamforming, when he desired user is he firs user, is T y () wx() 1 1 w T 1 aksk() k 1 T 1 1 1 1 1 s1() T 1 1 w a a w a a 1 If w 1 is designed such ha w a 1, a 1 and 1 a,and w1 1 T a 1, hen y 1 () = s 1 (), i.e., even if wo co-channel signals appear a he deskop compuer, only our desired signal is exraced. An inuiive explanaion of his resul is ha we design a weigh vecor for s 1 (), i.e. w 1, such ha he anenna array 1 forms nulls in all he DOA's excep 1. Similarly, w is designed such ha he paern has is nulls in all he esimaed DOA's excep 1. In his case, only he desired signal, i.e. s (), is exraced. 3. SIMULATION SETUP s () A block diagram of our simulaion is shown in Fig. 4. Firs, an endoscope image is convered o a binary sequence o be ransmied. Then he ransmied sequence is scrambled. The scrambled signals are sen o he encoder in order o perform he channel coding. The inerleaver inerleaves he encoded signals o reduce he correlaion beween he successive bis. The signals afer inerleaving are sen o he mapping and he pilo inserion blocks, where he mapping is used for mapping he signal o he QPSK consellaions diagram. The S/P converer convers he serial signal o parallel signals and IFFT ransforms he ransmied parallel signals o ime domain. Finally, he ime (4) 50 domain signals are convered o a serial signal by he P/S converer and padded wih guard inerval (GI). Fig. 4. A block diagram of our simulaion esbed. Then he signals are modulaed and ransmied. The ransmied signal can be represened as [5] N SD 1 s () wtsym () dk, n expj M( k) F ( TGI ) k 0 NST / pn 1 Pk expj kf( TGI), (5) knst / where w TSYM is he recangular window funcion having symbol-inerval widh; T G1 is he guard inerval; p n+1 is he cyclic exension of 17 sequence elemens which can deermine he polariy of he pilo subcarrier; P k and d k,n are he ransmied pilo signal and he daa signal, respecively; k and n are he indices of he OFDM symbol and subcarrier, respecively; F is he frequency spacing of adjacen subcarriers; M(k) is he index of he IFFT inpu; N SD and N ST are he numbers of he subcarrier of daa ransmission and pilo daa, respecively. The d k,n can be represened as d ) K, k, n ( I k, n jqk, n MOD (6) where K MOD is he normalized facor; I k,n and Q k,n are he inphase and quadraure componens of he signal, respecively. In his simulaion, he modulaion uilized in he simulaion is QPSK, and he ransmission daa rae used in his simulaion is 1 Mbps. Oher parameers of IEEE 80.11g are shown in Table II. The number of IEEE 80.11g mulipah signals is four. The 37

Biomed. Eng. Appl. Basis Commun. 006.18:46-54. Downloaded from www.worldscienific.com by 148.51.3.83 on 04/05/18. For personal use only. 51 Table II. Seup of relaive parameers Informaion daa rae 1 Mbi/s Modulaion QPSK OFDM Cener frequency.4 GHz Error correcing code K = 7 (64 saes) convoluional code Coding rae 1/ Number of subcarriers 5 OFDM symbol duraion 4.0 s Guard inerval 0.8 s Occupied bandwidh 16.6 MHz disribuion of DOA is Laplacian. The angle spread is 60 degree and he mean is 0 degree. Only he direc pah of main signal is considered. The number of anenna array elemen is eigh. Assume ha he spaial signaures will no change during he execuion of beamforming algorihms. The channel model proposed in [19] is adoped in his simulaion and shown in Fig. 5. The impulse response of he channel can be represened as c( ) l0 k0 exp( j ) ( T ) where T l is he arrival ime of he l-h cluser; is he k-h mulipah signal arriving a he receiver in he l-h cluser; is he phase of he received signals and is generaed by normal disribuion; is he power gain of he k-h mulipah signal in he l-h cluser. Fig. 5 shows ha he mean square value of power gain,, decreases when he oal and cluser power-delay ime consans, and, increase. This channel model is an exponenial decay. The ampliude of each mulipah signal is generaed from he Rayleigh disribuion o model mulipah fading. ( ) Fig. 5. Channel model used in our simulaion. (7) The variance depends on he delay spread of each mulipah signal. When he delay spread of he mulipah signal is longer, he ampliude is smaller. The variance of each mulipah signal can be represened as 1 exp( T / ) (8) s T exp( T / ) exp( / ) T 0 T 1... T l 0 RMS exp( kt / T ), k 0 k 0 s RMS l (9) Vol. 18 No. 5 Ocober 006 1 1 h (10) k N 0, k jn 0, k where Ts is he sampling ime; k is he index of he received mulipah signals; TRMS is he roo mean 1 square delay ime; and N(0, ) ) k is he Gaussian 1 disribuion wih mean and variance equal o 0 and k respecively; 0 is he reference variance o generae he variance of he k-h pah and k is he variance of he k-h pah. Possible values of TRMS are given in Table III [0]. In his simulaion, TRMS is chosen o be 100 ns. Table III. Indoor Values of TRMS [0] Environmen Delay Spread Home < 50 nsec Office ~ 100 nsec Manufacuring floor 00-300 nsec A he receiver, he anenna array receives mulipah and direc pah signals, and DDOA and PIDOA beamforming algorihms are uilized o generae he weighing vecor o form he radiaion paern for he desired signal. In his simulaion, assume ha he esimaed DOA'S of received signals are correc. Then, he guard inerval of received signal is removed. Nex, he S/P converer convers he serial signal o is parallel forma and FFT ransforms he received parallel signals o frequency domain, and he frequency domain signals are convered o a serial signal by he P/S converer, followed by running he inverse procedures of he ransmier such as pilos removing, demapping, deinerleaving, decoding and descrambling. Finally, we obain he binary sequence, conver he sequence ino is decimal represenaion, and reconsruc he image. 4. SIMULATION RESULTS AND DISCUSSION Generally, a 4-bi bimap image can be divided ino hree componens: R, G, B. In order o save simulaion ime in some cases, we conver an RGB image o is gray-level version and only he gray level image is used as he inpu o he proposed sysem in hose cases. A ypical capsule endoscope image is shown in Fig. 6. To verify ha he simulaion sysem performs correcly, we ake he gray-level capsule endoscope image and conver i o a binary sequence. The binary 38

Biomed. Eng. Appl. Basis Commun. 006.18:46-54. Downloaded from www.worldscienific.com by 148.51.3.83 on 04/05/18. For personal use only. BIOMEDICAL ENGINEERING- (a) (b) Fig. 6. A ypical capsule endoscope image. (a) Color image; (b) Gray-level image of (a). sequence is he inpu o he sysem and ransmied over an addiive whie Gaussian noise (AWGN) channel. Then, he receiver convers he binary sequence back o he corresponding decimal value and reconsrucs he ransmied image. Fig. 7(a) and 7(b) demonsrae he performance resuls in erms of he curves of BER (Bi Error Rae) v.s. SNR (Signal o Noise Raio) and PSNR (Peak Signal o Noise Raio) v.s. SNR, respecively, where PSNR is defined by I max PSNR 10log 10, MSE (11) r where Imax 1, and r is he number of bis o represen a pixel; MSE is he mean square error beween he image ransmied and he image received via he wireless channel. In he case considered here, Imax=55. Curve 0 and Curve 1 in Fig. 7(a) are he simulaion resuls for he heoreic QPSK signal and he IEEE 80.11g signal ransmied over AWGN channel, respecively. Noe ha he IEEE 80.11g signal is also QPSK modulaed and he channel coding is removed emporarily a his ime. The resuls show ha Curve 0 and Curve 1 are fied very well, meaning ha he simulaion plaform works correcly. Curve is he simulaion resul when IEEE 80.11g along wih a 1/ convoluional coder and Vierbi decoder is used. Comparing Curve wih Curve 1, we can find clearly ha he resul for IEEE 80.11g wih convoluional channel coding is much beer han ha wihou using i. Curve 3 in Fig. 7(a) is he simulaion resul of he IEEE 80.11g signal ransmied over a muliaph channel. Curve 3 and Curve in Fig. 7(a) demonsrae ha due o he guard inerval used in he IEEE 80.11g, he performance of IEEE 80.11g over he mulipah channel is almos as good as ha over he AWGN channel. Curve 4 and Curve 5 are he simulaion resuls of using he IEEE 80.11g wih DDOA and PIDOA beamforming algorihms, respecively. The performance improvemen wih a smar anenna sysem is very significan. Theoreically, he performance improvemen using PIDOA 5 beamforming should be beer han ha of using DDOA beamforming, bu our simulaion shows almos idenical resuls. Since a 1/ convoluional coder is used in he simulaion, mos bis in error are correced by channel coding. As a resul, he BER's for he DDOA case and for he PIDOA case are very close. However, uilizing channel coding will no enhance he performance of PIDOA beamforming algorihm, whereas i is useful for he DDOA case. For Curve 4 or 5, when he BER is 10-3, he corresponding SNR is abou -4.5 db. Thus, comparing Curve 4 or Curve 5 wih Curve 3, he smar anenna (a) (b) Fig. 7. Comparison of simulaion resuls in differen siuaions. (a) BER v.s. SNR; (b) PSNR v.s. SNR (heoreic: heoreic values for QPSK; c: wih coding; mp: mulipah; D: DDOA; PI: PIDOA). 39

Biomed. Eng. Appl. Basis Commun. 006.18:46-54. Downloaded from www.worldscienific.com by 148.51.3.83 on 04/05/18. For personal use only. 53 sysem can approximaely provide a 7.5 db anenna gain. Fig. 7(b) shows he PSNR performance for all he cases in Fig. 7(a) excep for Curve 0. As expeced, he PSNR value of he IEEE 80.11g wih he smar anenna sysem is he greaes because a lower BER generally implies a higher PSNR value for he same SNR value. Boh BER and PSNR are he objecive crieria o evaluae he performance of he IEEE 80.11g. A subjecive one is given nex. Fig. 8 shows he received images in differen cases when SNR is 1 db. Fig. 8(a) and Fig. 8(b) show he received images ha were ransmied over he AWGN and he muliaph channels, respecively. The image qualiy of hese images is obviously no good enough for normal diagnosis. Fig. 8(c) and Fig. 8(d) are he received images when he DDOA and he PIDOA beamforming algorihms are used, respecively. Obviously, image ransmission using IEEE 80.11g and he smar anenna sysem resuls in he image ha is almos idenical o he original one shown in Fig. 6(b). The proposed smar anenna sysem can significanly improve he qualiy of he received image. Fig. 9 shows he simulaion resuls of he color image ransmied over differen channels. The SNR in he simulaion is again se o 1 db. Fig. 9(a) and Fig. 9(b) are simulaion resuls of he color image ransmied over he AWGN and he mulipah channels, respecively. Obviously, he image ransmied over he mulipah channel is worse han ha ransmied over he AWGN channel. Fig. 9(c) and (a) (b) Vol. 18 No. 5 Ocober 006 Fig. 9(d) give simulaion resuls when he DDOA and he PIDOA beamforming echniques are exploied, respecively. Comparing Fig. 9(c) or Fig. 9(d) wih Fig. 9(b), i is no difficul o find ha he received image qualiy using beamforming is much beer han ha wihou using beamforming. I shows again ha he smar anenna can indeed improve he performance of he IEEE 80.11g in image ransmission applicaions. 6. CONCLUSION In a capsule endoscopy sysem, a paien swallows a capsule-like camera and inesine images aken by he camera are ransmied o a deskop compuer. The capsule endoscope images can help a docor o idenify he inesine-relaed problems. Bad image qualiy may inerfere he docor from making a normal diagnosis. This work proposes a smar anenna soluion o enhance he reliabiliy and efficiency of IEEE 80.11g for medical image ransmission. We apply Dominan DOA and Pseudo Inverse DOA beamforming algorihms o he IEEE 80.11g sysem. When he capsule endoscope image is ransmied over he proposed plaform, performance of IEEE 80.11g wihou smar anenna is worse han ha wih smar anenna. The simulaion resuls show ha he proposed smar anenna sysem can approximaely provide a 7.5 db anenna gain. The simulaion resuls in erms of BER v.s. SNR also demonsrae ha using (a) (b) (c) (d) Fig. 8. Simulaion resuls in differen siuaions (SNR = 1). Refer o (a) Curve ; (b) Curve 3; (c) Curve 4; (d) Curve 5 in Fig. 7. (c) (d) Fig. 9. Simulaion resuls for color image ransmission (SNR = 1) in differen condiions. Refer o he condiion used o obain (a) Curve ; (b) Curve 3; (c) Curve 4; (d) Curve 5 in Fig. 7. 40

Biomed. Eng. Appl. Basis Commun. 006.18:46-54. Downloaded from www.worldscienific.com by 148.51.3.83 on 04/05/18. For personal use only. BIOMEDICAL ENGINEERING- convoluional channel coding can improve he performance of he DDOA beamforming and make i as good as he PIDOA beamforming. We also use acual images o evaluae he performance of he proposed sysem. As a resul, based on he image ransmied over he mulipah channel using IEEE 80.11g wihou smar anenna, a docor may no have enough confidence o make a meaningful diagnosis, whereas here will be no diagnosis difficuly for he docor when he smar anenna sysem is exploied because he qualiy of he received image is well preserved. REFERENCES 1. Iddan G, Meron G, Glukhovsky A, and Swain P: Wireless capsule endoscopy. Naure 000; 405: 417.. Adler DG and Gousou CJ: Wireless capsule endoscopy. Hospial Physician 003; 39(5): 14-. 3. Chang FL and Miaou SG: Micro-reconnaissance machine inside he body-capsule endoscope. Science Developmen Monhly 005; 57-61. 4. IEEE Sd 80.11-1997 Informaion Technology - Telecommunicaions and Informaion Exchange Beween Sysems - Local and Meropolian Area Neworks - Specific Requiremens - Par 11: Wireless Lan Medium Access Conrol (MAC) and Physical Layer (PHY) Specificaions 1997; i-445. 5. IEEE Sd 80.11g - 003, IEEE Sandard for Informaion Technology - Telecommunicaions and Informaion Exchange Beween Sysems - Local and Meropolian Area Neworks - Specific Requiremens Par II: Wireless LAN Medium Access Conrol (MAC) and Physical Layer (PHY) Specificaions 003; i-67. 6. Zheng TB: In-Deph Analysis on 80.11 Wireless Nework Technique, DrSmar Press Co., Ld, Taiwan, 004. 7. Godara LC: Applicaions of anenna arrays o mobile communicaions, Par II: Beamforming and direcion-of-arrival consideraions. Proc. IEEE, 1997; 85: 1195-145. 8. Haykin S, Reilly J, Kezys V, and Veraschisch E: Some aspecs of array signal processing. Proc. Ins. Elec. Eng. F 199; 139(1): 1-6. 9. Swales SC, Beach MA, and McGeehan JP: The performance enhancemen of muli-beam adapive base saion anennas for cellular land mobile radio sysems. IEEE Trans. Vehicular Technology 1990; 39: 56-67. 10. Li YJ, Feuersein NJ, and Reudink DO: Performance evaluaion of a cellular base saion muli beam anenna. IEEE Trans. Vehicular Technology 1997; 46: 1-9. 54 11. Sano H, Murai H, and Miyake M: Performance of beam-combining scheme for DS-CDMA using weighing facor based on inerference level. Proc. VTC 99 1999; : 993-997. 1. Choi SW, Shim DH, and Sarkar TK: A comparison of racking beam arrays and swiching beam arrays operaing in a CDMA mobile communicaion channel. IEEE Anennas Propaga. Mag. 1999; 41: 10-. 13. Lin CS: Performance Evaluaion of Smar Anenna in WLAN IEEE 80.11a Sysem. Maser hesis, Elecrical Engineering, Naional Dong Hwa Universiy 004. 14. Roy R and Kailah T: ESPRIT - esimaion of signal parameers via roaional invariance echniques. IEEE Trans. Acousics, Speech and Signal Processing 1989; 37: 984-995. 15. Paulraj A, Roy R and Kailah T: A subspace roaion approach o signal parameer esimaion. Proc. IEEE 1986; 74(7): 1044-1045. 16. Jeng SS, Okamoo GT, Xu G, Lin HP, and Vogel WJ: Experimenal evaluaion of smar anenna sysem performance for wireless communicaions. IEEE Trans. Anennas and Propagaion 1998; 46(6): 749-757. 17. Jeng SS, Huang CY, Lai CY and Huang TJ: Performance evaluaion of DOA based beamforming in W-CDMA sysem. Proc. VTC 01 001; 4: 65-656. 18. Winers JH, Salz J and Gilin RD: The impac of anenna diversiy on he capaciy of wireless communicaion sysem. IEEE Trans. Commun. 1994; 4: 1740-1750. 19. Saleh A and Valenzuela R: A saisical model for indoor mulipah propagaion. IEEE Journal on Seleced Areas in Commun. 1987; 5(): 18-137. 0. O'Hara B. and Perick A: The IEEE 80.11 Handbook - A Designer's Companion, Sandards Informaion Nework, IEEE Press, 1999. 41