CLPC: Cross-Layer Product Code for Video Multicast over IEEE 802.11 Kihwan Choi and Sunghyun Choi School of Electrical Engineering and INMC, Seoul National University, Seoul, Korea E-mail:
[email protected],
[email protected] Abstract— In the wireless network, both random errors and burst errors do occur due to the channel noise. In order to combat such heterogeneous error patterns, we employ an adaptive error protection based on product code. The proposed product code-based protection in network includes cross-layer protection, because each layer experiences a different type of channel errors. We improve the decoding efficiency by adopting a proper protocol stack which allows upper-layers to exploit erroneous packets for the packet reconstruction. In addition, we consider the video multicast over IEEE 802.11 and formulate an aggregated multicast video-quality maximization problem under bandwidth constraint. The results show that the proposed protection scheme improves the aggregated multicast videoquality thanks to the cross-layer optimization. keywords: video multicast over IEEE 802.11, Reed-Solomon (RS) code, forward error correction (FEC), product code, crosslayer optimization.
I. I NTRODUCTION The use of IEEE 802.11 wireless local area networks (WLANs) as an extension to the existing wired IP networks is growing at a rapid pace. The high bandwidth provided by WLAN technologies such as IEEE 802.11a/g and the upcoming IEEE 802.11n will ultimately lead to their increasing use for multimedia networking. Recent multicast wireless video, such as digital TV streaming and interactive conference, poses several key requirements which need to be satisfied in order to provide a reliable and efficient transmission: 1) easy adaptability to bandwidth variations of wireless network; 2) relatively small delay bound; and 3) support for multiple receivers. In this paper, we investigate the robust and efficient multicast of video over WLANs. We specifically consider the highspeed WLAN standard, IEEE 802.11a [2], which offers high transmission rate up to 54 Mb/s, enabling the transmission of delay sensitive audio/visual (AV) traffic. This paper proposes a novel vertical system integration that enables the joint optimization of the various protection strategies existing in the protocol stack. In the remainder of this paper, we refer to this vertical system integration strategy as cross-layer protection. The error control strategies that can be implemented at the various layers, namely, 1) application-layer forward error correction (FEC), 2) media access control (MAC)-layer FEC, and 3) physical-layer (PHY) mode adaptation will be investigated for the efficient multicast of video over 802.11a. In addition to the proposed product code-based protection strategy, we consider video multicast over WLAN. As the demand on multicast digital video streaming grows fast, the high-speed low-cost WLANs are considered the link-layer protocol for the wireless multicast. In this paper, we formulate an aggregated multicast video-quality maximization
problem with bandwidth constraint, and develop a crosslayer adaptation problem in order to find the optimal protection strategy considering the given channel conditions and multicast policy. Moreover, we specifically consider the AP (access point)-initiated contention-free transmission period which IEEE 802.11e supports by means of TXOP (transmission opportunity) of HCCA (hybrid coordination function controlled channel access) [3], since the use of TXOP is an efficient scenario for real-time traffic transmission. Reed-Solomon (RS) product code-based protection without feedback can improve the error-recovery capacity when both burst and random errors exist. This is the reason why product code is one of popular protection schemes in the digital audio/video system of the optical storage [8]. In general, an error control scheme of the optical storage is composed of two parts, i.e., inner coding and outer coding. The inner decoder is intended for the correction of most of the small random byte errors and the detection of the larger burst errors. The outer decoder is intended for the correction of burst errors which the inner decoder could not correct. In this paper, we propose a cross-layer product code-based protection scheme in order to resist random and burst errors in both the WLAN and the wired-IP network. In order to recover from random errors, MAC-layer FEC adopts intra-packet RS coding. Even if MAC-layer error recovery fails, application-layer FEC based on inter-packet RS coding manages errors in the packets. The reason why we adopt FEC for both the MAC and application layer is clear in that automatic repeat request (ARQ)-based error recovery does not guarantee delay bound which is critical for real-time transmission, and that we consider a multicast environment where multiple receivers suffer from diverse channel errors. Today’s IEEE 802.11 do not support reliable multicast. It only provides an unreliable multicast protocol. With IEEE 802.11, packets with multicast MAC address can be transmitted to multiple receivers, but there is no feedback, i.e., acknowledgment (ACK) exchange between the sender and the receivers [1]. Therefore, the reliability for multicast over WLAN cannot be guaranteed. For more robust transmission, application-layer inter-packet RS coding can be used to reconstruct lost packets. Another benefit is that application-layer FEC does not introduce excessive delay while retransmission does. Generally speaking, the interactive-multimedia transmission is fragile to delay, but can tolerate some loss. However, feedback such as TCP ACK can introduce excessive delay since at least two end-to-end delays are necessary to exchange feedback and retransmitted packets. For the same reason, application-layer FEC is used for multimedia transmission over WLAN in [6].
Moreover, with ARQ-based recovery, it is a challenging problem to collect feedbacks from multicast receivers for each video packet. However, inter-packet RS coding can exploit the diversity of packet error patterns among the multicast receivers [13]. The adaptive cross-layer protection strategies proposed in this paper can be applied to any video coded bitstreams, such as non-scalable MPEG-4 or H.264 coded bitstreams, datapartitioned bitstreams, MPEG-2 and MPEG-4 hybrid spatiotemporal scalable bitstreams, fully scalable wavelet video bitstreams, etc. In this paper, however, we employ MPEG4 Fine-Granularity-Scalability (FGS) for the compression of the video data, because it can provide an easy adaptation to bandwidth variations and device characteristics [12]. There exist various studies on layered protection in the IEEE 802.11 and video transmission. In [9], a transmitter station can change its physical transmission rate for combating wireless channel errors. In [4], a MAC-layer FEC protection is adopted in order to reduce packet losses due to wireless channel errors. The problem of resilient realtime video streaming over IEEE 802.11b for both unicast and multicast transmission is considered in [11]. For the multicast, progressive video coding based on MPEG-4 FGS is combined with FEC. Similarly, the combination of MPEG-4 FGS with a scalable FEC and an unequal protection strategy are proposed in [12]. The cross-layer consideration necessary for wireless multimedia transmission is introduced in [6][7]. Our adaptive cross-layer protection strategy is pursued as follows. • We analytically derive the packet loss probability of the proposed scheme at various channel conditions for the given PHY mode, MAC-layer FEC, and applicationlayer FEC. • An analytical model is developed to characterize the end-to-end distortion of multicast video quality based on the packet loss probability and bandwidth constraint in WLAN. • Based on the end-to-end distortion model, a cross-layer protection strategy is developed to adapt such parameters as a) the video bit-rate, b) the application-layer FEC, c) the MAC-layer FEC, and d) PHY-layer modulation, in order to maximize the aggregated multicast video quality across multicast receivers. The rest of the paper is organized as follows. In Section II, we present the proposed cross-layer product codebased protection protocol. We then analyze the packet loss probability after decoding and the resultant video quality in Section III. In Section IV, we describe the cross-layer adaptation problem of determining the optimal protection strategy in order to maximize the aggregated multicast videoquality. In Section V, we present the numerical results, and the paper concludes in Section VI. II. P ROTOCOL D ESCRIPTION A multicast video streaming network in consideration is depicted in Fig. 1. The streaming server produces video
Fig. 1.
Fig. 2.
Network topology for multicast video streaming.
Protocol stack adopted for the cross-layer product code.
and parity packets (as shown in Fig. 4) transmitted as IP datagrams in the wired network. The IP datagrams arrive at the AP, which the multicast receivers are associated with. The AP encapsulates the received MAC service data units (MSDUs), i.e., IP datagrams, as the 802.11 MAC protocol data units (MPDUs or simply MAC frames) [1]. With the received MAC frames from the AP, the receivers reconstruct IP datagrams, and then video packets at the application layer. The protocol stack is depicted in Fig. 2. We adopt UDP-lite for transport-layer protocol in order to enable the application layer to exploit erroneous payloads from UDP-lite [14]. Similar to the UDP-lite, the network-layer protocol, i.e., IP, also checks errors only in the IP header part so that the application layer can access erroneous packets. Packets suffer from two different types of errors, i.e., packet loss and packet corruption. Since congestion in the wired network can cause buffer overflow at routers, packet loss can occur as packets are forwarded through a chain of wired links. Whereas, the wireless channel can cause both packet loss and packet corruption. The errors occurred in the MAC frame header disable the receiver to recognize the important information such as the destination address so that this error-patterns result in packet loss. On the other hand, the errors in the payload of the MAC frame lead to the corruption of the original data. In order to recover from random errors in the wireless network, we adopt intra-packet RS coding for the MAClayer FEC. The MAC-layer RS encoder at the sender of the wireless network adds redundancy to each original MAC frame. Then, the receiver(s) can recover from random errors
of received MAC frames by RS error-correction decoding. As shown in Fig. 3, one MAC frame protected by MAClayer RS coding consists of one header codeword and M independent payload codewords. The MAC header codeword uses (106, 70) RS code with the error-correction capability of 18 bytes, whereas each payload codeword uses (NM , KM ) RS code with the error-correction capability of tM (= (NM − KM )/2). All the RS codes in consideration are defined over GF(256), i.e., an RS code symbol is 8 byte long. The message block of the MAC header codeword, i.e., 70 bytes, is subdivided by MAC header and higher-layer headers, and includes 802.11e MAC header (26 bytes), LLC header (8 bytes), IP header (20 bytes), UDP header (8 bytes), and application-layer header (8 bytes). Therefore, MACheader recovery guarantees that there is no error in the headers of all the associated protocols. For the recovery from erasures, we adopt inter-packet RS coding for the application-layer FEC. As shown in Fig. 4, across multiple video packets, symbols along the same column produce symbols of the parity packets by using (NA , KA ) RS code. Then, the streaming server sends the parity packets along with the original video packets. For the application-layer FEC, erasure-correction decoding is used so that (NA − KA ) erased symbols can be recovered. The decoding mechanism at the receiver(s) works as follows: 1) The receiver checks errors of the received MAC frame using Frame Check Sequence (FCS) based on Cyclic Redundancy Check (CRC)-32 at the tail of the MAC frame. If the CRC fails, the receiver tries to reconstruct the header using a fixed, e.g., (106, 70) MAC header RS code. If both CRC and MAC-header decoding fail, the received MAC frame is discarded. 2) If the MAC header decoding of an erroneouslyreceived frame succeeds, each payload codeword is decoded with (NM , KM ) RS code, where the code information is described in the application-layer header. If a payload codeword decoding fails, i.e., due to over tM byte errors, the MAC layer declares “erasure” for the codeword. 3) The MAC layer reconstructs the MSDU using both successfully recovered codewords and erasures. The reconstructed MSDUs are forwarded to the application layer through LLC/IP/UDP-lite layers. Because IP and UDP-lite check only the errors in their headers, even partially erased packets are forwarded onto the application layer. The MAC layer also forwards the erasure information of the partially erased packets onto the application layer. 4) The application layer recovers the original video packets from the partially-erased packets and successfullyreconstructed packets by using inter-packet RS erasurecorrection decoding. All the video packets can be reconstructed as long as the number of correctly-received symbols from the MAC layer is greater than or equal
Fig. 3.
MAC frame format for MAC-layer RS error-correction code.
Fig. 4.
Application-layer RS coding across packets.
to KA across all the columns. III. M ATHEMATICAL A NALYSIS In this section, we derive video quality with bottom-up approach. First, the bit error probability of the physical layer given a channel condition is derived. Based on the result, the decoding-error probabilities of the MAC header and payload codewords after RS error-correction decoding are derived. Then, the overall packet loss probability after applicationlayer RS erasure-correction decoding is also derived. Finally, we describe the video quality based on the packet loss probability and the wireless bandwidth constraint. A. 802.11a PHY Bit Error Probability The analysis of the 802.11a PHY error performance here is based on those given in [9]. In this analysis, it is assumed that the noise over the wireless medium is additive Gaussian noise (AWGN) channel. According to the analysis the bit error probability, Pb , after the Viterbi decoding is bounded by the value depending on the selected PHY mode m and the average SNR, γ. Note that a PHY mode basically represents a specific transmission rate of the 802.11a. The probability PP LCP of error in the PLCP header including both SIGNAL (of 24 bits) and SERVICES (with 7 scrambler initialization bits) fields, when PHY mode m is used for the SERVICE field of the PPDU (PHY protocol data unit), can be determined by PP LCP = 1 − (1 − Pe (1, 24)) · (1 − Pe (m, 7)), where the probability Pe (m, l) of error in a block of length l bits, assuming random bit errors with Pb of PHY mode m, is determined by Pe (m, l) = 1 − (1 − Pb )l .
More details of the analysis on the 802.11a PHY can be found in [9]. B. Cross-Layer Product Code Performance
codewords. Hence, the conditional probabilities for the event, Ds , of the successful reconstruction of the given packet are expressed as
We here assume that each packet is received with the independently and identically distributed average SNR, γ, where the probability distribution function is given by f (γ). Then, the symbol (of one byte) error probability, when PHY mode m is used, is given by Ps (m) = Pe (m, 8), where Ps (m) is the MAC-layer symbol error probability. In the following, we use the notation Ps instead of Ps (m) for simplicity as long as there is no confusion. First, we derive the probabilities which MAC-layer RS error-correction decoding leads to. We assume that MAClayer RS coding for a header codeword uses (NH , KH ) RS code with the error-correction capability tH = (NH − KH )/2. Note that NH = 106 and KH = 70 from Fig. 3. Given γ and the packet loss probability, Pwired , at the wired network, the success probability of RS error-correction decoding for the MAC header codeword can be expressed as P [Hs |Γ = γ] =
tH i=0
P [Ds |Hsc , L = l] M l l = , P [Bs |Hs ]i (1 − P [Bs |Hs ])l−i i=KA i
(6)
where Hsc is the complement of the event, Hs . From the above equations, the overall packetreconstruction success probability of the product code can be obtained by P [Ds ] = P [Hs , Q = 0] +
M N A −1
× P [Hs , Q = q]P [L = l] +
N A −1
P [Ds |Hsc , L = l]P [Hsc ]P [L = l], (7)
l=KA
(1)
where Bs denotes the event of successful MAC-layer RS error-correction decoding for a given MAC payload codeword. Then, given f (γ), P [Bs |Γ = γ] · P [Hs |Γ = γ] f (γ)dγ, (3) P [Bs |Hs ] = P [Hs ] P [Hs |Γ = γ] · f (γ)dγ.
P [Ds |Hs , Q = q, L = l]
q=1 l=KA
NH N −i Psi (1 − Ps ) H i
where Hs denotes the event of successful RS error-correction decoding for the MAC header codeword. Similarly, the success probability of RS error-correction decoding for each MAC payload codeword can be expressed as tM NM N −i (2) P [Bs |Γ = γ] = Psi (1 − Ps ) M , i i=0
P [Hs ] =
(5)
× (1 − PP LCP ) · (1 − Pwired ),
where
P [D Q = q, L = l] q s |Hs , l l i l−i , = P [Bs |Hs ] (1 − P [Bs |Hs ]) i=KA i
(4)
Now, we derive the probabilities which application-layer RS erasure-correction decoding leads to. As shown in Fig. 4, each byte of video packets is used as the message symbol to produce redundancy bytes in parity packets. Given a packet, we define a random variable, Q, as the number of the payload codewords in the packet with unsuccessful MAC-layer RS error-correction decoding, and another random variable, L, as the number of packets, belonging to the same application-layer RS erasure-correction decoding group (i.e., of NA video plus parity packets) with the given packet, with successful MAC-layer decoding for their header
where
NA − 1 P [L = l] = P [Hs ]l (1 − P [Hs ])NA −l−1 , l P [Hs , Q = q] = P [Hs , Q = q|Γ = γ]f (γ)dγ,
(8) (9)
and P [Hs , Q = q|Γ = γ] M = P [Bs |Γ = γ]M −q (1 − P [Bs |Γ = γ])q q × P [Hs |Γ = γ].
(10)
Then, the overall packet loss probability, P [Dsc ], is obtained by P [Dsc ] = 1 − P [Ds ], (11) where Dsc is the complement of the event, Ds . C. MPEG-4 FGS Video Coding for Wireless Transmission The adopted model of video quality in this section follows R-D model of FGS video [6]. The peak signal-to-noise 2552 ) is used as a measure of ratio (PSNR = 10 log10 MSE video quality, where MSE is the distortion per pixel in mean-squared-error. The PSNR of FGS video at the receiver with tolerable base-layer packet loss increases approximately linearly with the effective received bit-rate, REL,d , of the enhancement layer at the receiver after channel losses as follows PSNR = θ · REL,d + PSNRBL , (12)
where θ is the R-D model parameter, which depends on the spatio-temporal characteristics of the video sequence [12], and PSNRBL is the video quality (PSNR) of the base layer. Based on the very simple error concealment method [6], we can determine statistically the bit-rate of the video data that is received without errors. For the enhancement layer, a single packet loss within an enhancement-layer video frame causes the remainder packets associated with that video frame useless. Assuming Nf enhancement-layer packets are sent for the current video frame, and the video frame rate of fr (frames/s), the effective enhancement-layer bit-rate REL,d at the decoder given a packet loss probability, PEL , of the enhancement-layer packets (assuming equal error protection among enhancement sublayers) is REL,d =
Nf
(1 − PEL )i−1 · PEL · (i − 1)
i=1
+ (1 − PEL )
Nf
· Nf fr · 8KM · M. (13)
For the packet transmission, a fixed amount of HCCA TXOP is used by the AP every beacon interval. If we assume that HCCA TXOP without backoff is used, Nf can be computed based on bandwidth constraint as T XOPEL + SIF S 1 KA 1 · · , (14) Nf = · TEL + SIF S NA TB fr where T XOPEL is the length of HCCA TXOP for the enhancement-layer packet transmission, SIF S is short interframe space defined by the 802.11, TEL is the transmission time for an enhancement-layer packet depending on PHY mode and (NM , KM ), and TB is the 802.11 beacon interval [1]. Because we can exploit the adaptation parameters (i.e., NA , KA , NM , KM , and PHY mode, m) for a given channel condition and HCCA TXOP, our optimization problem can be further understood as to maximize the aggregated PSNR of multicast receivers for a given bandwidth constraint (i.e., HCCA TXOP) on the wireless channel. We delve into crosslayer optimization for video multicast over IEEE 802.11 in the next section. IV. C ROSS -L AYER O PTIMIZATION In this section, we formulate the optimization problem under wireless bandwidth constraint considering each layereddesign.
B. MAC-layer FEC At the receiver, PHY forwards the received PSDU (PHY service data unit) to MAC, and then MAC checks errors of the MPDU. With the IEEE 802.11 standard, MAC checks errors using CRC. If no error occurs, MAC forwards the MSDU to the upper-layer, while MAC discards the MSDU if the CRC fails. However, the intra-packet FEC for IEEE 802.11 WLAN is beneficial rather than CRC alone [4]. We consider the code adaptation of MAC-layer RS coding. C. Application-layer FEC and Video Rate Adaptation With a bandwidth constraint for video multicast, the maximum number of transmitted packets is determined by the given packet size and PHY rate. Given the maximum number of transmitted packets, the number of video packets and the number of parity packets should be determined considering the given packet loss probability in order to maximize video quality at the receivers. From Eq. (14), we can also adapt the application-layer RS code, the packet size, and the video frame rate. D. Bandwidth Constraint In wireless networks, it is typical that bandwidth constraint on multimedia exists for co-existing flows. One of the solutions for the co-existing problem is the admission control for each permitted flows considering the existing traffic flows. In IEEE 802.11e, admission control is done by asserting TXOP to each flow. In this context, TXOP limit of the video multicast can be interpreted as its bandwidth constraint. Given a TXOP limit, the flow is able to use the wireless channel during the allowed TXOP limit. Therefore, for an FGS video flow, the relation between TXOP for base and enhancement layers is given by T XOPlimit ≥ T XOPBL + T XOPEL ,
(15)
where T XOPBL denotes the TXOP length for the base-layer packet transmissions. Given base-layer bit-rate, RBL , and protection strategy, the necessary TXOP for the base-layer transmission, T XOPBL , can be expressed as RBL · TB NA · (TBL + SIF S), (16) T XOPBL = · 8KM · M KA where RBL is the encoding base-layer bit-rate, and TBL is the transmission time for a base-layer packet, which depends on NM , KM , and PHY mode. E. Multicast Optimization
A. PHY Mode Selection IEEE 802.11 WLAN is able to select the appropriate physical-layer (PHY) constellation and convolution code rate for a channel condition [9]. When the wireless channel suffers from severe noise, more robust PHY mode can overcome the channel noise and enable to exchange packets successfully.
Multicast service can vary due to the service policy. In this paper, we exemplify a multicast service model exploiting the proposed protection scheme. We consider the multicast receivers with heterogeneous channel conditions. The multicast policy of this paper is that all of the receivers should have a small distortion for the base-layer video, and the maximum aggregated video quality for the enhancement-layer video.
The employed video decoding system assumes that if a higher priority packet is lost (i.e., a base-layer packet or a packet containing a more significant enhancement layer bitplane), then the lower priority packets in the same video frame are discarded. Consequently, the packet loss rate of the base layer should be kept very small. In [10], the performance of non-scalable MPEG-4 base layers has been determined for a variety of channel conditions, and it has been determined that for most sequences, if the base-layer packet-loss rate PBL is lower than 1%, the overall FGS performance remains unaffected. Therefore, given a particular channel condition, a fixed RBL , and the target PBL , we can determine the error protection strategy in order to minimize T XOPBL while keeping PBL lower than 1%. If X denotes the set of all possible vectors of the adaptation parameters, i.e., (NA , KA , NM , KM , m), and Ptarget denotes the target packet loss probability of the base-layer packets, then we can define the adaptation-parameter selection problem for the base-layer video transmission as x∗BL = arg min T XOPBL x∈X
subject to
T XOPBL ≤ T XOPlimit , PBL,i ≤ Ptarget ∀i,
(17)
where PBL,i denotes the base-layer packet loss probability of receiver i out of multicast receivers. Thus, given the obtained T XOPBL , the optimal solution of a weighted summation problem can be constructed as wi PSNRi x∗EL = arg max x∈X
subject to
i
T XOPEL ≤ T XOPlimit − T XOPBL . (18)
where PSNRi denotes the average PSNR value of station i given SNR distribution, and wi is the given weight factor for receiver i depending on the multicast policy, where i wi = 1. In summary, based on the cross-layer product code-based protection, we can allocate the given bandwidth in order to maintain tolerable base-layer distortion across all the multicast receivers as well as to maximize the aggregated enhancement-layer performance for the multicast receivers. V. N UMERICAL E VALUATION In this section, we comparatively evaluate the proposed cross-layer product code-based protection scheme via numerical results based on the analysis made in previous sections. We assume that all the FGS video packets are transmitted via 802.11a PHY with 8 different PHY modes supporting 6 (mode 1), 9, 12, 18, 24, 36, 48, and 54 (mode 8) Mb/s. The fixed parameters values used for the evaluation are summarized in Table I. For simplicity, we assume that all base-layer video packets are correctly received, and all multicast receivers have an identical channel condition. Based on the assumption and the fixed parameters, we reduce the cross-layer optimization problem to finding the optimal KA in {3,9,15,21,27,33,39,45,51,57,63}, NM
TABLE I PARAMETER SETTING Parameter θ PSNRBL TB T XOPEL
Fig. 5.
Value 2.49 (dB/Mb/s) 20.29 (dB) 100 (ms) 30 (ms)
Parameter NA KM M Pwired
Value 63 200 5 0
Two-state discrete time Markov chain for the wireless channel.
in {200,206,212,218,224,230,236,242,248,254}, and m in {1,2,3,4,5,6,7,8} to maximize video quality. We consider two channel models, i.e., AWGN and timevarying wireless channels. Fig. 5 shows the two-state discrete time Markov chain modeling the time variation of the wireless channel. The wireless channel could be in either good or bad state. When the wireless channel is in good state, the corresponding SNR at each time instant is taken from a uniform distribution in the range of 15 to 30 dB, and when the wireless channel is in bad state, the SNR value is drawn from the range of 0 to 15 dB. The time spent in the good and bad are take from exponential distributions with rates 1/μg and 1/μb , respectively. Therefore, the state b and transition probabilities tg,b and tb,g equal to µgµ+µ b µg , respectively. Different values of t correspond to g,b µg +µb different wireless channel variation patterns. For example, if tg,b is close to 0 (1), the wireless channel tends to stay in good(bad) state for most of the time. We assume that the playout delay bound is large enough to reduce the jitter effects [5]. Large playout delay bound also enables interleaving between packets, exploiting efficient FEC code with large codeword length, and reducing the effect of burst channel errors. Therefore, we can assume the large and fixed codeword length for application-layer RS code, and random packet error with average packet loss probability. As shown in Fig. 6, the cross-layer product code-based protection (labeled as “CLPC”) leads to higher multicast video quality than the application-layer FEC protection with link adaptation (labeled as “AFEC+LA”) under AWGN channel. Here, “link adaptation” refers to the PHY mode selection. This is because MAC-layer FEC exploits RS errorcorrection decoding for the intra-packet error recovery, while application-layer FEC adopts the inter-packet error recovery based on RS erasure-correction decoding. Given an identical bit error probability for each packet, the intra-packet error recovery is more efficient than the inter-packet error recovery. Therefore, the proposed cross-layer protection does not exploit application-layer RS coding under AWGN channel. From Table II, we observe that the optimal KA is 63 for all five SNR values, and knowing that NA = 63, we confirm
TABLE II O PTIMAL PARAMETER SELECTION UNDER AWGN CHANNEL SNR (dB) 5 10 15 20 25
KA 63 63 63 63 63
NM 218 206 212 206 200
PHY Mode 3 4 6 7 8
TABLE III O PTIMAL PARAMETER SELECTION UNDER THE TIME - VARYING CHANNEL
Fig. 6.
Comparison under AWGN channel.
tg,b 0 0.1 0.2 0.3 0.4 0.5
KA 63 51 45 45 45 51
NM 206 206 218 206 206 230
PHY Mode 6 6 6 4 4 3
R EFERENCES
Fig. 7.
Comparison under the time-varying channel.
that the proposed scheme operates as the MAC-layer FEC protection without the application-layer FEC. Fig. 7 shows the video quality performance of the crosslayer product code (CLPC), application-layer FEC with link-adaptation (AFEC+LA), MAC-layer FEC with linkadaptation (MFEC+LA), and link-adaption alone (LA) over the time-varying wireless channel. We observe that the proposed CLPC consistently outperforms all other schemes across different tg,b values. If the wireless channel is in good state and does not change to bad state, the cross-layer product code-based protection mainly exploits the intra-packet RS coding. However, as the state-transition probability, tg,b , increases, it depends on the inter-packet RS coding rather than the inter-packet RS coding. Therefore, under the time-varying wireless channel, the proposed scheme utilizes both MAClayer and application-layer FEC. The adaptation parameters can be selected considering the variation of channel as shown in Table III. VI. C ONCLUSION In this paper, we propose a cross-layer product codedbased protection which improves video quality under bandwidth constraint. We also investigate multicast scenarios of FGS video streaming over WLAN and formulate the aggregated video-quality maximization problem. The analysis and numerical results show that the proposed protection scheme is more efficient than the other existing protection schemes considering both AWGN and time-varying channels.
[1] IEEE Std. 802.11-1999, Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications, Aug. 1999. [2] IEEE 802.11a, Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications, High-speed Physical Layer in the 5 GHz Band, Sep. 1999. [3] IEEE 802.11e, Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications: Medium Access Control (MAC) Enhancements for Quality of Service (QoS), Nov. 2005. [4] S. Choi, Y. Choi, and I. Lee, “IEEE 802.11 MAC-Level FEC with Retransmission Combining,” IEEE Trans. on Wireless Communications, vol. 5, no. 1, pp. 203-211, Jan. 2006. [5] J. F. Kurose and K. W. Ross, Computer Networking: A Top-Down Approach Featuring the Internet, 3rd Ed., Addison Wesley, 2004. [6] M. van der Schaar, S. Krishnamachari, S. Choi, and X. Xu, “Adaptive Cross-Layer Protection Strategies for Robust Scalable Video Transmission over 802.11 WLANs,” IEEE Journal of Selected Areas in Communications (JSAC), vol. 21, no. 10, pp. 1752-1763, Dec. 2003. [7] M. van der Schaar, N. S. Shankar, “Cross-Layer Wireless Multimedia Transmission: Challenges, Priciples, and New Paradigm,” IEEE Wireless Communications Magazine, vol. 12, no. 4, pp. 50-58, Aug. 2005. [8] S. Lin and D. J. Costello, Error Control Coding, Fundamentals and Applications, 2nd Ed., Englewood Cliffs, NJ, Prentice-Hall, 2004. [9] D. Qiao, S. Choi, and K. G. Shin, “Goodput Analysis and Link Adaptation for IEEE 802.11a Wireless LANs,” IEEE Trnas. on Mobile Computing (TMC), vol. 1, no. 4, pp. 278-292, Oct.-Dec. 2002. [10] M. van der Schaar and H. Radha, “Unequal packet loss resilience for fine-granular-scalability video,” in IEEE Trans. on Multimedia, vol. 3, no. 4, pp. 381-394, Dec. 2001. [11] A. Majumdar, D. Sachs, I. Kozintsev, K. Ramchandran, and M. Yeung, “Multicast and unicast real-time video streaming over wireless LANs,” in IEEE Trans. on Circuits Syst. Video Technol., vol. 12, no. 6, pp. 524-534, June 2002. [12] H. Radha, M. van der Schaar, and Y. Chen, “The MPEG-4 fine-grained scalable video coding method for multimedia streaming over IP,” in IEEE Trans. on Multimedia, vol. 3, no. 1, pp. 53-68, Mar. 2001. [13] J. Nonnenmacher, E. Biersack, and D. Towsley, “Parity-Based Loss Recovery for Reliable Multicast Transmission ,” in IEEE/ACM Trans. on Networking (TON), vol. 6, no. 4, pp. 349-361, Aug. 1998. [14] L. Larzon, M. Degermark, S. Pink, L. Jonsson, and G. Fairhurst, “The Lightweight User Datagram Protocol (UDP-Lite),” RFC 3828, July 2004.
