Performance Analysis of MPEG-4 Video Stream with FEC Error Recovery over IEEE DCF WLANs

Transcription

1 erormance Analyss o MEG-4 Vdeo Stream wth FEC Error ecovery over EEE DCF WLANs Cheng-Han Ln 1, Hua-Wen Zhang 2, Ce-Kuen Sheh 3 Department o Electrcal Engneerng Natonal Cheng Kung Unversty Tanan, Tawan, OC jhln5@hpds.ee.ncku.edu.tw 1 q @mal.ncku.edu.tw 2 sheh@ee.ncku.edu.tw 3 Wen-Shyang Hwang 4, * Department o Electrcal Engneerng Natonal Kaohsung Unversty o Appled Scences Kaohsung, Tawan, OC wshwang@mal.ee.kuas.edu.tw 4, * Abstract Due to the ncreasng use o moble and handheld devces, the perormance o vdeo streamng over wreless networks has emerged as an mportant concern n recent years. However, whle many analytcal models have been proposed or analyzng the system perormance, these models do not take explct account o the eects o error recovery. As a result, they al to provde an accurate ndcaton o the true vdeo qualty at the recever end. Accordngly, ths paper proposes a model or evaluatng the perormance o MEG-4 vdeo streamng wth Forward Error Correcton (FEC) over EEE Dstrbuted Coordnaton Functon (DCF) Wreless Local Area Networks (WLANs). The proposed model consders not only the eects o congeston and wreless rame losses, but also the FEC error recovery perormance n mprovng the perceved vdeo qualty at the recever end. The valdty o the proposed model s demonstrated by comparng the predcted values o the layable Frame ate (F) wth the results obtaned rom NS-2 smulatons and two exstng analytcal models, respectvely. The results conrm that the proposed model provdes an accurate ndcaton o the perceved qualty o FEC-protected MEG-4 vdeo streamng over DCF WLANs. Keywords- FEC; DCF; vdeo stream; WLAN. NTODUCTON Due to the convenence o wreless access, the use o moble computers and smart phones has ncreased sgncantly n recent years. Furthermore, wth the ncreasng coverage o Wreless Local Area Networks (WLANs), and the avalablty o ever greater data rates and bandwdth, the use o such devces to access nternet-based vdeo streamng applcatons has become partcularly common. Thus, the perormance analyss o vdeo streamng over wreless networks has emerged as an mportant ssue n the multmeda communcatons eld. The EEE standard [1] lays down two transmsson unctons or meda access over wreless channels, namely Dstrbuted Coordnaton Functon (DCF) and ont Coordnaton Functon (CF). n DCF, the deault transmsson uncton, the actve wreless nodes compete or channel access and any contentons among them are resolved usng the nary Exponental acko (E) method. y contrast, n CF, the communcatons wthn the network are coordnated by an Access ont (A) n such a way as to acheve a collson-ree servce. n the DCF mode, the rames sent by an actve staton may al to reach ther destnaton due to collsons wth the rames o other statons contendng or the same channel. Moreover, attenuaton, adng, scatterng and ntererence rom other actve sources may also prevent a rame rom reachng ts destnaton. n other words, rame losses may occur as a result o both congeston and wreless lnk errors. FEC (Forward Error Correcton) [2-6] s a wdely used technque or recoverng rame losses n wreless networks. n FEC schemes, the sender mtgates the eects o packet losses by transmttng redundant packets together wth the source packets such that a block can stll be reconstructed at the recever end even some o the packets wthn the block are lost durng transmsson. The lterature contans many models based on a twodmensonal Markov chan or analyzng the perormance o EEE DCF networks [7-10]. One o the rst models was presented by anch [7]. For smplcty, the model assumed that the number o retransmssons o a lost rame s unlmted,.e., the sender contnues to retransmt lost rames untl they are successully receved. y contrast, n the model proposed n [8], the rame retransmsson lmt prescrbed n EEE DCF s taken nto consderaton,.e., the rames are dropped by the sender collsons occur repeatedly wthn the maxmum retransmsson perod. However, the eects o wreless bt errors on the rame loss are gnored. Dong et al. [9] developed an enhanced model or obtanng accurate predctons o the EEE DCF perormance over lossy channels. However, the rame retransmsson lmt was gnored. N et al. [10] presented a model n whch the eects o both the rame retransmsson lmt and wreless bt errors were taken nto account. n general, the analytcal models presented n [7-10] ocus on the system perormance, but do not enable the vdeo qualty over EEE DCF WLANs to be drectly assessed. n most prevous studes on the qualty o MEG-4 vdeo streamng over wreless networks [11-13], the vdeo qualty was evaluated usng only network-level parameters, e.g., the packet loss rate or the packet delay. n [12-13], Zvan and Ke proposed a new perormance metrc, desgnated as the *Correspondence to: W.-S. Hwang s wth the Department o Electrcal Engneerng, Natonal Kaohsung Unversty o Appled Scences, Tawan.O.C. (emal: wshwang@mal.ee.kuas.edu.tw)

2 Decodable Frame ate (DF), or analyzng the vdeo qualty o MEG-4 vdeo streamng over WLANs. However, the assumptons regardng the wreless transmsson were overly smple (e.g., an average packet loss rate across the network). Ln et al. [14] proposed a more realstc model n whch the eects on the rame losses o wreless channel errors and transmsson collsons were both taken nto account. However, as wth the models presented n [12-13], the eects o error recovery on the MEG vdeo streamng qualty were not consdered. Ths paper proposes an analytcal model or evaluatng the perormance o MEG-4 vdeo streamng over EEE DCF WLANs wth FEC error protecton. n evaluatng the vdeo qualty, the model consders both congeston losses and wreless channel losses. n addton, the model takes account o the FEC error recovery perormance n mprovng the perceved vdeo qualty at the recever end. Fnally, the model enorces the rame retransmsson constrant prescrbed n EEE DCF. The valdty o the proposed model s vered by comparng the analytcal results or the layable Frame ate (F) wth those obtaned va NS-2 smulatons and two exstng analytcal models, respectvely. The remander o ths paper s organzed as ollows. Secton revews the background o the proposed model, ncludng the EEE DCF transmsson mode, the FEC error correcton scheme and the MEG-4 vdeo structure. Secton ntroduces the proposed perormance evaluaton model. Secton V demonstrates the valdty o the proposed model. Fnally, Secton V presents some bre concludng remarks and ndcates the ntended drecton o uture research.. ACKGOUND A. EEE DCF n EEE DCF, each staton competes or channel access based on the channel state and contenton wndow sze. the channel s busy, the staton senses the channel condton contnuously. the channel remans dle or a perod o tme equal to the Dstrbuted nter-frame Space (DFS), the staton ntalzes a backo tmer and deers transmsson or a randomly-selected backo nterval rom the contenton wndow sze. The backo tmer s decremented by one or every sensed dle tme slot and s rozen when the channel s sensed to be busy. The backo tmer s reactvated when the channel s sensed to reman dle once agan or more than one DFS perod, and the staton attempts to transmt a packet when the backo tmer reaches zero. n DCF, the contenton wndow sze s determned usng an exponental backo scheme known as nary Exponental acko (E). n E, the contenton wndow s doubled each tme a staton experences a transmsson collson, but s reset to ts mnmum value (CW mn ) whenever a successul transmsson occurs. When ntatng the transmsson o a new packet, the staton randomly assgns the backo tmer n the nterval o [0, CW mn - 1]. n accordance wth the dscussons above, the varaton o the DCF contenton wndow (CW) sze can be represented as ollows: CW Fgure 1. FEC protocol. Fgure 2. MEG GO structure, GO (2, 6). 2 CWmn CWmn collson successul where s the total number o aled transmssons o a packet. Followng each aled transmsson, CW s doubled untl a maxmum value o CW max = 2 m x CW mn s reached, where m s the maxmum number o retransmsson attempts. Once CW reaches CW max, t remans at ths value untl t s reset to CW mn as a result o ether a successul packet transmsson or the packet beng dropped due to the packet retransmsson lmt.. Forward Error Correcton (FEC) The FEC protecton mechansm enables the recever to correct errors / losses n the receved data wthout the need or any urther nteracton wth the sender. The basc prncple o FEC entals transmttng redundant rames (h) n addton to the source packets (k). The FEC approach s shown schematcally n Fg. 1. As shown, k rames o source data are coded at the sender together wth h redundant rames n such a way as to produce a total o n rames o coded data. Thus, provded that any k rames (source or redundant) are receved at the recever end, the source data can be successully reconstructed. Snce FEC schemes enable the recovery o source data rames whch would otherwse be lost, the eectve loss rate n the transmsson network s lower than the actual loss rate, and thus the perceved qualty o the vdeo stream at the recever end s mproved. n FEC codecs, the redundant packets are derved rom the orgnal packets usng conventonal codng theory technques. O the varous error correctng codes avalable, eed-solomon (S) code [15-16] has attracted partcular nterest snce ts use s explctly recommended n the nternet Engneerng Task Force (ETF) eal-tme Transport rotocol (T). Accordngly, the present study also assumes the use o S code n developng the proposed analytcal model.

3 C. MEG -4 Vdeo Structure n MEG-4 codng, the vdeo sequence s decomposed nto a seres o consecutve smaller unts known as a Group o ctures (GO). Each GO contans three types o rame arranged n a perodc sequence, namely ntra-coded (), redctve (), and drectonal (). -rames are smply stll mages wthn the sequence whch are ndependently encoded wthout reerence to any other rames. Meanwhle, -rames are orward predcted based on the relatve normaton provded by the prevous -rame or -rame. Fnally, -rames are encoded based on the relatve normaton provded by both the precedng - or -rame and the ollowng - or -rame. Fg. 2 shows a typcal GO structure wthn an MEG-4 vdeo sequence. Let N represent the number o -rames n the GO, N represent the number o -rames n the GO, and N represent the dstance between the - (or -) rame to the next -rame. Any GO structure can thereore be unquely dented by the notaton GO (N, N ). For example, the structure shown n Fg. 2 s desgnated as GO (2, 6). n other words, the rame sequence s as ollows:. n perormng vdeo streamng over wreless networks, the encoded vdeo rames are segmented nto small MAC data rames n accordance wth the maxmum transmsson unt (MTU) o the network. O all the rames receved at the recever end, some rames may not be decodable. n general, two derent types o undecodable vdeo rame exst: (1) drect undecodable rames,.e., an nsucent number o MAC rames belongng to the vdeo rame are receved to decode the rame; and (2) ndrect undecodable rames,.e., the loss o a dependent rame prevents the decodng o the receved rame.. OOSED ANALYTCAL MODEL Ths secton descrbes the analytcal model proposed n ths study or estmatng the perceved qualty o MEG-4 vdeo streamng over EEE DCF WLANs wth FEC error protecton. The secton commences by analyzng the eects o transmsson losses (e.g., collson losses and wreless losses) on the perormance o DCF WLANs. The proposed analytcal model s then ormally ntroduced and derved. A. erormance Analyss o EEE DCF WLANs n EEE DCF, the loss o a transmsson rame can be caused by two actors: () collson loss: resultng rom channel access contenton; and () wreless loss: resultng rom wreless ntererence. For each transmsson attempt, the alure probablty ( ) o a rame s constant and ndependent, rrespectve o the number o retransmssons. Thus, the probablty o rame transmsson alure can be expressed as 1 C E C E C E 1 1 where C and E represent the rame loss probabltes due to collson and wreless errors, respectvely. (Note that all o the notatons used n the proposed analytcal model are descrbed n Table.) Assumng that the data rame corrupton and ACK rame corrupton events are ndependent, t ollows that E E _ data E _ ACK E _ data, E _ ACK (1), (2) where E_data and E_ACK are the Frame Loss robabltes (FLs) o the data rame and ACK rame, respectvely. Assumng that the bt errors are unormly dstrbuted over the whole rame, E_data and E_ACK can be expressed respectvely as E 1 _ data 1 E 1 _ ACK 1 data _ length E _ bt ACK _ length E _ bt, (3), (4) where E_bt s the t Error robablty (E) o the wreless channel. n DCF, a rame transmsson attempt s made only when the backo tmer alls to zero. However, collsons may occur multple actve statons commence transmsson smultaneously. n calculatng the collson loss rate, t s assumed that each rame colldes wth a constant and ndependent probablty, C. Under steady-state condtons, each staton transmts a rame wth probablty. Consequently, the collson probablty or any staton competng or channel access s equal to C 1 n 1 1. (5) Combnng Eq. (1) and Eq. (5), the probablty o a rame transmsson alure s obtaned as 1 n E. (6) n the analytcal model proposed n ths study, the maxmum backo stage (m) s assumed to be equal to the maxmum retransmsson tme. Hence, a rame wll be dropped when the contenton wndow s at ts maxmum sze and a loss occurs as the result o ether a collson event or wreless errors. From the model presented n [10], the probablty o an actve staton transmttng n a randomly-chosen tme slot s gven by m m1 m1 1 CW mn. (7) Equatons (6) and (7) represent a nonlnear system wth two unknown varables, and. n Eq. (1), denotes the alure probablty each tme a rame transmsson s attempted over the DCF WLAN. However, rame retransmsson ncreases the probablty o a partcular rame beng successully receved. n developng the proposed analytcal model, an assumpton s made that each rame can be transmtted a maxmum o T max tmes beore beng dscarded by the sender. The eectve alure probablty o each rame s thereore gven by T Tmax max eectve C 1 Tmax C 0. (8). Analytcal Model or MEG-4 Vdeo Streamng wth FEC Error ecovery n practcal FEC-protected DCF WLANs, the redundant data are derved rom the orgnal source data usng a varety o e

4 derent error correctng codes. As descrbed n Secton., t s assumed n ths study, that the redundant data are derved usng eed-solomon (-S) erasure code [15-16]. Gven the use o -S code at the FEC codec and an eectve rame loss probablty o eectve, the probablty o a successul rame transmsson s gven as ollows [17]: n 1 k n n, k, eectve eectve eectve n, (9) where denotes all possble combnatons o the rames successully receved wthn a block, rrespectve o whether these rames are orgnal source rames or FEC-generated redundant rames. Accordng to the MEG vdeo structure shown n Fg. 2, the successul transmsson probabltes o the -, - and -rames n the GO are gven respectvely as S S S S S S, S, eectve, S,, S, eectve eectve, (10) where S, S and S are the number o MAC -, - and - rames, respectvely; whle S, S and S are the number o redundant -, - and -rames, respectvely. The authors n [17] proposed a perormance metrc desgnated as the layable Frame ate (F) or evaluatng the qualty o vdeo streamng over lossy networks. The F s dened as the rato o the expected number o decodable vdeo rames at the recever to the total number o vdeo rames transmtted by the sender. (Note that the F s computed on a per-second bass.) The eectve GO transmsson rate can be computed as G F 1 N N, (11) where N and N are the number o - and -rames n the GO, respectvely, and F s the encodng rame rate per second. n MEG-4 vdeo streamng, the -rames are always decodable provded that they are successully transmtted snce they are encoded ndependently o any o the other rames n the GO. As a result, the F or -rames s equal to the number o -rames transmtted successully over the network. n other words, the F s gven smply as G. (12) -rames are decodable only they are successully transmtted and the dependent precedng - or -rame s also decodable. Thus, the F or -rames s gven by N 1 n N 1 2 N. (13) TALE. E C E_data E_ACK E_bt T max eectve,, S, S, S S, S,S N, N F G,, NOTATONS USED N THE OOSED ANALYTCAL MODEL robablty o MAC rame transmsson alure. robablty o MAC rame loss due to wreless error. robablty o MAC rame loss due to collson. robablty o MAC data rame loss. robablty o ACK rame loss. t Error robablty (E). Maxmum number o retransmssons o MAC rame Eectve loss rate o MAC rame gven maxmum number o retransmssons equal to T max robabltes o successul -, -, and -rame transmsson. Numbers o MAC -, -, and -rames. Numbers o FEC redundant -, -, and - rames. Numbers o - and -rames n GO. Encodng rame rate per second. Eectve GO transmsson rate. layable rame rates o -, -, and -rames over entre vdeo sequence. Overall playable rame rate o FEC-protected MEG vdeo stream. As a result, the F or all the -rames n the entre vdeo sequence s equal to N G 1 1 N 1. (14) -rames are decodable only they are successully transmtted and the dependent precedng and succeedng - or -rames are both decodable. Consecutve -rames depend on the same - and -rames and thereore have the same F. Consequently, consecutve rames can be treated as a sngle group. The nal group depends on both the precedng - rame and the succeedng -rame. Thereore, the F o the - rames s gven by j j 1 when when 0 N 1 N. (15) Consequently, the F o all the -rames n the vdeo sequence s equal to N N 0 0 N G 1 N 1 N. (16) Clearly, the overall F or a FEC-rotected MEG vdeo stream s equal to the sum o the Fs o the -, - and - rames, respectvely. That s, G 1 1 N 1 N 1 N 1 N. (17)

5 TALE. AAMETE SETTNGS [7]. acket payload (L data ) 8184 bts Slot tme 50 s ACK (L ACK ) 240 bts DFS 128s MAC header 272 bts SFS 28s HY header 128 bts CW mn 32 Channel Data ate 1 Mbps T max 5 E ( E_bt ) 10-5 C 3.91 N 3 C 2.05 N 8 C 1.52 Fgure 4. Varaton o F wth number o actve statons gven poor channel condton (E = 10-4 ). Fgure 3. Smulaton topology. V. NUMECAL ESULTS n ths secton, the proposed analytcal model s valdated by comparng the predcted results or the F wth the analytcal results obtaned rom the models presented n [8, 9] and the NS-2 smulaton results presented by [18]. The consdered network topology s shown n Fg. 3. n comparng the F results o the varous methods, the number o wreless nodes wthn the network was vared n the range o 5 to 50. As descrbed n Secton, the model proposed n ths study takes account o both the retransmsson lmtaton and the eects o channel errors. y contrast, the models presented n [8, 9] consder only one o these eatures but not both. Moreover, the proposed model consders the F perormance o a vdeo stream delvered over a FEC-protected WLAN, whereas the models n [8, 9] gnore the eects o FEC error recovery. The parameter settngs used n the analytcal models and smulatons are shown n Table. n modelng the network perormance, t was assumed that the wreless statons operated under saturaton condtons,.e., each staton always had a packet avalable or transmsson. Furthermore, an assumpton was made that vdeo trac was sent rom a randomly selected node and receved by a randomly chosen destnaton node. Meanwhle, the other nodes n the network were assumed to transmt UD lows at a constant bt rate (C). The experments were perormed usng the "Mr. ean" vdeo trace [19]. The vdeo sequence comprsed rames (.e., rames, rames, and rames) wth a GO structure o (N =3, N =8). n transmttng the rames, the maxmum packet sze was set to 1000 bytes. Fgure 5. Varaton o F wth number o actve statons gven good channel condton (E = 10-6 ). A. Model Valdaton As the number o nodes n the network ncreases, the number o collsons also ncreases. The greater number o collsons causes more packets to be lost, and thus the F reduces, as shown n Fgs. 4 ~ 7. t s noted n Fgs. 4 ~ 7 that the F predctons obtaned usng the model proposed n ths study are n better agreement wth the smulaton results than those obtaned rom the analytcal models presented n [8,9]. For a bad channel condton (E = 10-4, Fg. 4), most o the rame losses are the result o wreless losses. n the model proposed n [8], channel errors are not consdered, and thus the F s overstated. Furthermore, snce channel errors are gnored, the model predcts an dentcal F perormance rrespectve o the channel condton (see Fgs. 4 and 5 correspondng to E = 10-4 and E = 10-6, respectvely). n the model presented n [9], the number o retransmssons s assumed to be unlmted, and thus the F s overstated or both channel condtons. n the case o a good channel condton (E = 10-6, Fg. 5), most o the rame losses are the result o channel contenton. Due to the very low error rate, the F predctons o the current model and the model presented n [8], respectvely, are very smlar. However, n general, the results presented n Fgs. 4 and 5 show that the proposed model provdes a more relable and robust evaluaton o the perceved qualty o FECprotected MEG-4 vdeo streamng over DCF WLANs

6 Fgure 6. Varaton o F wth number o actve statons gven m = 4 rame retransmsson opportuntes. Fgure 8. Varaton o F wth FEC overhead as uncton o E. (Node = 20, m=5) Fgure 7. Varaton o F wth number o actve statons gven m = 6 rame retransmsson opportuntes. under typcal (.e., varyng) wreless channel condtons. As shown n Fgs. 6 and 7, the F ncreases (.e., the perceved vdeo qualty mproves) as the maxmum number o retransmssons (m) s ncreased. The F results obtaned rom the proposed model are n good agreement wth the smulaton results or both values o m. For m = 4, the analytcal models presented n [8, 9] both overestmate the F (see Fg. 6). For a gven number o statons, the F predcton o the model n [9] s hgher than that o the model n [8] due to the assumpton o an unlmted number o retransmssons. However, gven a greater number o retransmsson opportuntes (e.g., m = 6), all o the statons are able to transmt most o ther packets successully. Thereore, as shown n Fg. 7, the F predctons o the two analytcal models are very smlar; even under heavy network load condtons.. erormance Analyss o Vdeo Stream gven FEC Error ecovery n ths sub-secton, the proposed analytcal model s used to evaluate the eect o the FEC error recovery mechansm on the perceved MEG-4 vdeo streamng qualty n EEE DCF WLANs under varous network condtons. Fg. 8 shows the varaton o the F wth the FEC overhead as a uncton o the bt error rate (E). Note that the FEC overhead s dened as the rato o the number o FEC redundant rames to Fgure 9. Varaton o F wth number o statons as uncton o FEC overhead. (E=10-5, m=5) the total number o rames (source plus FEC redundant). n other words, a 20% FEC overhead mples that one redundant rame s transmtted wth every our source rames. As expected, or all values o the E, the vdeo qualty mproves wth an ncreasng FEC overhead due to the greater successul decodng probablty. However, gven a poor channel condton (E = 10-4 ), the vdeo qualty s serously degraded even when a large number o redundant packets are njected nto the transmsson stream snce many packets (both source and redundant) are lost durng the transmsson process. Fg. 9 shows the eect o the FEC overhead on the perceved qualty o the vdeo stream under varous network loads. Under lght network loads (.e., n 10), the vdeo qualty s relatvely unaected by the number o redundant rames added to the source rames. However, gven a larger number o actve statons, the collson probablty ncreases, and thus a hgher FEC overhead results n a lower eectve rame loss rate and an mproved vdeo qualty. V. CONCLUSON Ths paper has proposed an analytcal model or evaluatng the vdeo qualty o MEG-4 vdeo streamng over FECprotected EEE DCF WLANs. The proposed model consders not only the eects o congeston and wreless rame losses, but also the perormance o the FEC error recovery

7 mechansm n mprovng the perceved vdeo qualty at the recever end. The valdty o the proposed model has been conrmed by comparng the predcted results or the layable Frame ate (F) wth the results obtaned rom NS-2 smulatons and two exstng analytcal models, respectvely. The results have conrmed the ablty o the proposed model to yeld an accurate predcton o the perceved qualty o MEG- 4 vdeo streamng over DCF WLANs wth FEC protecton. n a uture study, the proposed model wll be extended to the case o channel ntererence [20-21] over EEE e [22-23] based networks wth os support. ACKNOWLEDGMENT Ths research was supported by the Natonal Scence Councl, NSC E MY3 and NSC E MY3. EFEENCES [1] EEE standard or Wreless LAN Medum Access Control (MAC) and hyscal Layer (HY) speccatons, EEE Standard , June [2] M. Elaoud and. amanathan, Adaptve use o error-correctng codes or real-tme communcaton n wreless networks, n EEE Conerence on Computer and Communcatons, NFOCOM1998, vol. 2, pp , Apr [3] H. Seeroglu, Y. Altunbasak, O.Gurbuz, and O. Ercetn, ate dstorton optmzed jont A-FEC scheme or real-tme wreless multmeda, n EEE Conerence on Computer and Communcatons, CC2005, vol. 2, pp , May [4] C. H. Ln, C. H. Ke, C. K. Sheh, and N. Chlamkurt, An enhanced adaptve FEC mechansm or vdeo delvery over wreless networks, n nternatonal Conerence on Networkng and Servces, CNS2006, [5] C. H. Ln, C. K. Sheh, N. Chlamkurt, C. H. Ke, and W. S. Hwang, A ED-FEC mechansm or vdeo transmsson over WLANs, EEE Transactons on roadcastng, vol. 54, no. 3, pp , Sep [6] L. Han, S. ark, S. Kang, and H.. n, An adaptve Cross-Layer FEC mechansm or vdeo transmsson over WLANs, n nternatonal Conerence on nternet, C 2009, pp , Dec [7] G. anch, erormance analyss o the EEE dstrbuton coordnaton uncton, EEE Journal o Select Areas Communcaton, vol. 18, no. 3, pp , [8] H. Wu, Y. eng, K. Long, S. Cheng, and J. Ma, erormance o relable transport protocol over EEE wreless LAN: analyss and enhancement, n roc. EEE NFOCOM, New York, U.S., June 2002, pp [9] X. J. Dong and. Varaya, Saturaton throughput analyss o EEE wreless LANs or a lossy channel, EEE Communcaton Letters, vol. 9, no. 2, pp , [10]. N, T. L, T. Turlett, and Y. Xao, Saturaton throughput analyss o error-prone wreless networks, Wley Journal o Wreless Communcatons and Moble Computng, vol. 5, no. 8, pp , [11] J. Shn, J. Km, and C. C. Kuo, ualty o servce mappng mechansm or packet vdeo n derentated servces network, EEE Transactons on Multmeda, vol. 3, no. 2, pp , [12] A. Zvan,. E. Wolnger, J. F. ezende, O. C. Duarte, and S. Fdda, Jont adopton o os schemes or MEG streams, Multmeda Tools and Applcatons, vol 26, no. 1, pp.59 80, [13] C. H. Ke, C. H. Ln, C. K. Sheh, W. S. Hwang, and A. Zvan, Evaluaton o streamng MEG vdeo over wreless channels, Journal o Moble Multmeda, vol. 2, no. 1, pp.47 64, [14] C. H. Ln, C. H. Ke, C. K. Sheh, and N. Chlamkurt, The packet loss eect on MGE vdeo transmsson n wreless networks, n EEE nternatonal Conerence on Advanced normaton Networkng and Applcatons, ANA2006, Venna, Austra, Aprl 2006, pp [15] V. Guruswam and M. Sudan, mproved decodng o eed-solomon and algebrac-geometrc codes, EEE Transactons on normaton Theory, vol. 45, pp , [16] V. oca, Desgn, evaluaton and comparson o our large block FEC codecs, LDC, LDGM, LDGM starcase and LDGM trangle, plus a eed-solomon small block FEC codec, NA res. rep , Jun [17] H. Wu, M. Claypool, and. Knck, Adjustng orward error correcton wth temporal scalng or TC-rendly streamng MEG, ACM Transactons on Multmeda Computng, Communcatons, and Applcatons, vol. 1, no. 4, pp , [18] NS-2 smulator, [19] Vdeo trace, [20] Y. C. Ln, C. H. Ln, W. S. Hwang, and C. K. Sheh, FOCA Channel Assgnment Strategy Usng Four artally Overlappng Channels n WMNs, EEE Asa acc Wreless Communcaton Symposum, AWCS 2011, Sngapore, Aug 22-23, [21] C. H. Ln, N. Chlamkurt, S. Zeadally, and C. K. Sheh, erormance Modelng o MEG-4 Vdeo Streamng Over EEE Usng Dstrbuton Coordnaton Functon, Wreless Communcaton and Moble Computng, vol. 11, no. 9, pp , Sept [22] C. H. Ln, C. K. Sheh, C. H. Ke, N. Chlamkurt, and S. Zeadally, An adaptve cross-layer mappng algorthm or MEG-4 vdeo transmsson over EEE e WLAN, Telecommuncaton Systems Journal, vol. 42, no. 3, pp , Dec [23] C. H. Ln, Y. C. Wang, C. K. Sheh, and W. S. Hwang, An Unequal Error rotecton Mechansm or Vdeo Streamng over EEE e WLANs, Computer Networks, vol. 56, no. 11, pp , July 2012.