An error recovery transmission mechanism for mobile multimedia

Transcription

1 An error recovery transmission mechanism for moile multimedia Min Chen *, Ang i School of Computer Science and Technology, Hunan Institute of Technology, Hengyang, China Astract This paper presents a channel coding transmission mechanism in wireless and moile networks y using the cross-packet transmission method of the su-lock structure to solve the prolem of the poor quality of the network moile multimedia. The present channel coding technology has the protective effect aiming to the data transmission in the moile network. However, the present channel coding technology faces many kinds of network transmission phenomenon under the limited andwidth constraint, including those comprehensive prolems caused y the network congestion and the signal error. The poor efficiency of the data recovery and the andwidth utilization is still one of the channel coding s main defects. Compared with the present technologies, the experimental results demonstrate that the proposed method can achieve etter andwidth utilization y means of using su-lock coding structure to comat with small amount of transmission errors. Furthermore, the data recovery performance can e improved ased on the two following facts aout the cross -packet transmission: ) the channel coding capacity is eyond a single packet and 2) oth packet loss and signal error can e recovered simultaneously. Keywords: cross-packet, channel coding, forward error correction,, error recovery, moile multimedia Introduction With the increasing of the Internet popularizing rate, the source of the digital contents presents a prosperous development tendency through the wireless access technology. According to the statistics from the ministry of economic affairs digital content industry office, the output in the moile application service attained 2.86 illion dollars in 22 and its annual growth rate is 3.45%. Adding to the emerging digital converge service, the future moile users can receive several of data from the telecommunication, Internet and television. The investigating reports in the Google moile network and the user s ehavior also show that 9 percent of the smart phone users watch videos through the moile phones, and 3 percent of users use the video function once everyday. The traditional multimedia data transmission model easily causes the data loss under the influence of communication quality and the andwidth. For example, the lack of the andwidth under the network heterogeneity causes the packet loss and the poor efficiency of the communication environment causes the data error. The results make users can not otain the complete data so that the correct information can not e showed. The present network does not possess the quality of service mechanism, the channel coding technology and the low latency method can provide the reliale data transmission. The typical channel coding applications include the wireless communication protocol such as 82. Wireless AN and 82.6 WiMAX, and the video streaming service such as the video conference. The core technology of the channel coding is forward error correction. The original data is coded to the data in the transmission terminal and the two data are transmitted together. The receiving terminal conducts the decoding action to recover the original data within the tolerant data error numers in terms of the condition of the data reception. Although the utilization can offer the error recovery function, the cost of adding the andwidth should e paid off. When the coding data cannot recover the network transmission error, the prolem of wasting the andwidth can e caused. According to the advantages and the disadvantages analysis of the network transmission in the application, the high efficient mechanism in the data transmission system should e designed to improve the service quality in the moile network video in terms of the andwidth resources whose wireless moile network is relative precious. The present known channel code has two types: the signal packet model and the cross packet model. The Figure shows that the signal packet model merges the original data and the data into a signal transmission packet, and its advantage is that the decoding procedure can e conducted after the receiving terminal gets a packet. Therefore, it time delaying is low. While its disadvantage is that it just can deal with the data loss caused y the signal error. What s worse, the whole packet loss has no protective aility and the efficiency of the channel coding can e limited y the length of the maximum transmit unit in the network, namely, the size of the data numer in EFC is limited. When the condition of the signal error is severe, the data protective aility provided y the signal packet model will e lowed. The cross packet model regards the packet in the coding & decoding as unit and the * Corresponding author s chenminhit@63.com 54

2 original packet can codes packet so that the prolems of the packet loss and the data loss caused y the signal error can e solved simultaneously, its structure is shown as the Figure 2. The coding & decoding data numers in the are very large and the receiving terminal needs to wait for time to collect all transmitting packets, the time delaying is rather high. In addition, using the larger data numer packet to recover the signal error whose data numers are relative small in the andwidth utilization is also not economic. Under the limited transmission andwidth, the andwidth cost in the cross packet model has low economic enefit so that its protective aility is limited. The two different kinds of the models have advantages and disadvantages. In general, the signal packet model is fit for the wireless communication protocol to maintain a certain communication quality and the cross packet model is widely applied to all kinds of the network application service. With the variety of the network. The channel coding must adapt to the changing of all network environments and the accumulation of the transmission conditions. For example, the numers of the data loss may accumulate the final results formed y the different transmission factors. Therefore, no matter whether the effect of the signal packet model is not enough to handle the packet loss and the MTU limitations or low andwidth utilization in the cross packet model, all of these prolems are the needed improved ojects. source packe t it errors packe t Packet loss FIGURE The packet structure in the signal packet model source packet packet su locks and the coding su locks to form the transmission packets, namely, two kinds of the su locks can e assemled into the transmission packet individually. The two different assemling methods allow the coding su locks to cross many transmission packets. When the mechanism connects to the wireless moile network transmission system, the parameter and the formula can e set in terms of the target application s network environment. Moreover, the calculation of the su lock size, the numers of the coding su locks and the assemling of the transmission packet can e done so that the proposed mechanism can have the optimal efficiency. 2 The error recovery mechanism with the su lock structure Aiming to the signal packet model s and the cross packet model s efficiency limitation in the moile data transmission, the error recovery mechanism with the su lock structure is proposed to adapt to the coding unit in the wireless transmission environment, and the data s network transmission is conducted y the cross packet model. As shown in the Figure 3, the original data packet can e disassemled into a smaller source su lock and it regards coding as the unit, and the needed coding su locks can e produced according to the present network transmission conditions. Then the su locks can e assemled into the transmission packet, and the coding su locks are allowed to cross many transmission packets during the assemling procedure. On the one hand, the mechanism can effectively improve the utilization of the andwidth. On the other hand, the packet assemling method in the cross packet can avoid the MTU limitation and has a higher elasticity in the utilization. In order to make the su lock structure apply into the transmission system, the mechanism further design the needed data packet format and formulate the data transmission procedure. The related details are shown in the following su chapters. it errors Packet loss Source packet packet FIGURE 2 The packet structure in the cross packet model The coding mechanism with the su lock structure y connecting the wireless moile network transmission system is connected is proposed to minimum the coding data numers for protect the most common it errors and the packet loss so that the efficiency of the andwidth utilization and the efficiency of the data recovery can e considered simultaneously. The proposed mechanism in the operation can e divided into two parts:. the original data packet can e disassemled into a smaller source su lock and it regards coding as the unit, and the needed coding su locks can e produced according to the present network transmission conditions. 2. The su locks can e assemled into the transmission packet, and the assemling method can merge the source it errors Packet loss FIGURE 3 The cross packet model with the su lock structure 2. DATA PACKET FORMAT The length of the su lock is allowed to e changed so that how to define the positions of some individual su locks for correctly conducting the E\ coding & decoding procedure in a packet is the designing key point. Take the applied layer as an example, the Figure 4 illustrates that the data format in the transmission packet is applied with the use of the su lock coding structure. In 55

3 general, the s error recovery technology is applied in the multimedia data network transmission. The kind of the transmission uses UDP (User Datagram Protocol) as the transmission layer protocol. And then connects with the application layer protocol in the RTP (Real-time Transmission Protocol) to control and supervise the data s transmission conditions. In order to carry all kinds of the data flow patterns, RTP has formulated the s control header to conduct the data s coding & decoding procedure. The packet can e divided into header fields and data fields with the present multimedia data transmission structure. The header fields include data header, su lock header and header. The data header in the Figure 4 is compatile with the standard of the RTP/UDP protocol, and the header is compatile with the related standards of the RFC59 protocol. Code Amount Blocks (code=) Data header Su-lock header Checksum... Checksum n Data header Su-lock Source Data ( data) FIGURE 4 The packet format in the su lock... Su-lock n The su lock header can include the content source code segment which identify the composition source of the su locks in the packet, the amount segment which record and store the n numers of the su locks in the packet, multi validation and checksum segments which individually record the validation and the value in the ~n su lock. The validation and the value can e used to check whether the su lock can happen errors. The data fields put multi source su lock information or the su lock information. There are three packet kinds through the su lock coding in terms of the difference of the packet contents:. source data, 2. data and 3. Mixed source data with the data. The former two packets can e regarded as the transmission packet assemled y the source su lock and the su lock individually. Code segment can identify the packet types. For example, the Code segment is composed y two its: the first it is the source data it. When the it is remarked as I, the su lock in the data field includes the source su lock; the second it is data it. When the it is remarked as I, the su lock in the data field includes the su lock. Therefore, when the Code segment is, the transmission packet is the source data packet; while the Code segment is, the transmission packet is the data packet. When the Code segment is, the transmission packet is assemled y the source data packet and the data packet. The Code segment remains unused under the definition. When the Code segment is, the Amount segment needs a new extensional locks segment for remarking the numers of the data packet. original data su lock control unit coder packet encapsula tion unit network decapsula tion unit su lock assemly unit Feedack message original data decoder FIGURE 5 System structure and transmission procedure The display order of the su locks in the data field is further defined under the model of the mixture source su locks and the su locks, that is, the source su lock exists efore the su lock. Therefore, the former Amount Blocks in the transmission packet are source su lock and the later Blocks are the su lock. When the Code segment is not, the Blocks segment does not exist. The design considers that the mixture packet condition just appears in the convergence of the source su locks and the su locks. Therefore, the Blocks segment can e regarded as the need utilization for the effective use of the andwidth. The Figure 5 is the mechanism s structure and the operational procedure in the data transmission system, including six system units:. Su lock control unit: According to the recovery information in the su lock assemly unit, the size of the su lock can e decided and the numers of the su lock can e calculated. 2. EC coder: Conduct the coding procedure in the su lock. 3. Packet encapsulation unit: Encapsulate the su lock into the transmission packet. 4. De capsulation unit: De capsulate the transmission packet into su lock. 5. Su lock assemly unit: Filter the error su lock, collect the complete receiving su locks and recover the statistical information of the receiving data to the su lock control unit. 6. decoder: Conduct the decoding procedure and recover the original data. Compared with the traditional data transmission system, the mechanism adds the new specific units: Su lock control unit and Su lock assemly unit. The system can set the parameter and the formats in terms of the target application s network environment so that the method can otain the optimal effect in all network transmission environments. The main calculation is to determine the length of the su lock, the numer of the su lock and the needed su lock numers. The su lock length and the su lock numer are calculated 56

4 y the su lock control unit in reference to the su lock assemly unit, while the packet encapsulation unit and the de capsulation unit are according to the former descried packet format. In order to recover the packet loss caused y the network congestion, the coder and decoder can adopt to the common Reed-Solomon (RS) code. As to the RS code, when the su lock happens error, it can e regarded as the su lock loss and comine with the cross packet s transmission model. Therefore, the data recovery in the transmission error and the packet loss can e handled simultaneously. lock is k, the expected recovery rate in the source su lock is R. The following Equation can calculate the h numer in the su lock and its otained total su lock numer is (k+h). R k h ( k h k h i i )( P total ) ( P total ). (6) i k i If the k numer of the source su lock is 8, the packet total loss rate Ptotal is.2 and the expected data recovery rate R is.95, the minimun value of the h numer su lock is 5 in terms of the Equation (6). 3 The optimal calculation of the su lock The optimal calculation method is introduced and the performance verification can e conducted. If the current length of the su lock is it and the su lock error rate caused y the it errors is P, the it error rate is P. The Equation () is as follows: P ( P ) P P B B In the Equation (), the su lock error rate PB can e otained y the recovery information in the su lock assemly unit. According to the it error rate is P, the optimal su lock length can e calculated (opt) and the present su lock length can e updated. If the needed data header length in the su lock is d it, the transmission efficiency in the su lock can e defined as E: ( ) ( ) E P d d (). (2) The should e differential. If the differential result is set as, the opt can e otained as shown in the Equation (3) opt d In P d In P d In P 2 In( P ) 2 2 ( ) 4 ( ) ( ). (3) The new su lock error rate PB_new can e otained with the use of the su lock length opt. ater. The su lock error rate PB_new caused y the newly otained su lock length P can e predicted. P _ ( P) opt B new. (4) If the PB_new adds the packet loss rate PC caused y the packet loss in the su locks, the total loss rate Ptotal in all su locks can e otained. P P P. (5) total B _ new c In the meantime, the packet loss rate PC can e otained from the recovery information in the su lock assemly unit. The recovery condition in the source data is that the numers of the total su locks received from the coder is larger than or equal to the k numer in the source su lock. If the numer of the present source su 4. Experimental result and efficient analysis The validations of the experimental environment are as follows: The packet loss rate of the network is set as %, 2% and 3%, the loss numer of the average packet is set as 3, and the it error rate is set as -, -2, -3, -4, -5 and -6. According to the corresponding network from the %, 2% and 3% packet loss rate, the utilized andwidth is 224 KB/s, 45 KB/s, KB/s respectively. The fixed packet length is ytes, the original data source can e compressed into Foreman video whose format is MPEG4 QCIF and its frame per second is 3. The comparative oject in the experiment is the traditional cross packet model. The size and the numer of the su lock in the mechanism is determined y the aove optimal method discussed in the former chapters. In the setting expected recovery rate aspect, the unequal error protection mechanism is adopted in terms of the method in the reference []. The important data can e distriuted a large numer of packet and the unimportant data can e distriuted a small numer of packet. The negative effects of the data recovery caused y the limited andwidth can e reduced to the minimum. The efficient parameter adapts to the playale frame rate, and its definition is that if the present playale frame numer per second id f, and the playale frame rate is (f/3). Therefore, the higher the playale frame rate is, the more the source data in the video has, namely, the recovery affect in the etter. The testing video in the experiment in compressed through the 3 frame numers per second defined y the National Television Standards Committee, the conversion of the playale frame rate is.. The four kinds of the experiments are conducted successively. The former three mainly analyzes the difference of the playale frame rate y oserving the %, 2% and 3% packet loss rate. The Figure 6-8 present the three experimental results. The forth experiment counts the total utilization of the andwidth otained from the former three experiments whose experimental result is as shown in the Figure 9. The result in the first experiment is as shown in the Figure 6 whose packet loss rate is %. The higher playale frame rate under the condition of the transmission it error which is larger than -3 can e otained in the paper. The - it error usually represents the worst transmission environment so that nearly all 57

5 playael frame rate COMPUTER MODEING & NEW TECHNOOGIES 24 8(2B) transmission packet can happen errors. What s worse, the efficient difference can not e oserved. The result in the second experiment is as shown in the Figure 7 whose packet loss rate is 2%. The higher playale frame rate under the condition of the transmission it error which is larger than -3 can e otained in the paper. Compared with the first experiment, the packet loss rate in the transmission environment can e increased and the utilized andwidth can e reduced. The efficiency provided y the paper under the condition of the -2 it error rate can e improved largely, in which the playale frame rate is.8 and the frame numer per second is 25. In addition, the playale frame rate is. in the traditional cross packet model and its frame numer per second is just 3. The result in the third experiment is as shown in the Figure 8 whose packet loss rate is 3%. The higher playale frame rate under the condition of the transmission it error which is larger than -4 can e otained in the paper. The magnitude of the improvement can e increased with the increasing of the it error rates. The efficiency parameter in the forth experiment is data overhead whose definition is as follows: Overhead is equal to the packet header data size in all transmission packets plus to the data sizes of all su locks. The lower the Overhead s, the higher the efficiency of using the andwidth y the has. The Figure 9 shows the differences of the data usage etween the two comparative methods. The data sources are from the first experiment to the third experiment, and the packet loss rate is respectively the data usage under the %, 2% and 3% packet loss rate. Overhead(kiloytes) playael frame rate Bit error rate FIGURE 8 the third experimental result: the packet loss rate is 3% Congestion packet loss rate (%) FIGURE 9 the overhead statistical result from the first experiment to the third experiment Compared with the traditional packet model, the experiment results show that the data usage in the mechanism is lower. In fact, the mechanism can possessed etter transmission efficiency in the former experimental results. In addition, the mechanism can adopt the timesaving umer to finish the source data recovery so that the optimal efficiency under the limited andwidth can e otained. playale framerate FIGURE 6 The first experimental result: the packet loss rate is % Bit error rate FIGURE 7 The second experimental result: the packet loss rate is 2% 5 Conclusion Aiming to the limited andwidth, the transmission error and the network congestion suffered from the multimedia data transmission in the wireless moile network, the cross packet transmission mechanism with the su lock channel coding structure. Comining with the asic concepts etween the su lock structure and the cross packet transmission, the data packet format in the application layer is further formatted and the transmission system structure and the optimal calculation in the su lock can e illustrated so that the mechanism in the implement interface is complete. After the comparison and analysis of the related experiments, the mainly efficiency in the mechanism can e applied in the time-saving data sizes for recovering the error data so that the data recovery efficiency in the limited andwidth can e largely improved. The advantage makes the mechanism e applied in the multimedia application whose andwidth is limited in the moile network. 58

6 References [] Availale: [2] ser-ehavior-survey [3] ocal and metropolitan area networks Specific requirements Part : Wireless AN medium access control (MAC) and physical layer (PHY) specifications Amendment 8: medium access control (MAC) quality of Service enhancements IEEE Standard [4] Part 6: Air interface for fixed roadand wireless access systems IEEE Standard 82.6, Oct 24 [5] Rizzo 997 Effective erasure codes for reliale computer communication protocols ACM SIGCOMM Computer Communications 27(2) [6] Choi S, Choi Y, ee I 26 IEEE 82. MAC-level scheme with retransmission comining IEEE Trans on Wireless Communications 5 2 [7] Hwang S-h, Hwang S-O, Myung S, Yang H-K, Park K-M 23 Method for generating forward error correction packet in multimedia system and method and apparatus for transmitting and receiving forward error correction packet US Patent (issued) US A [8] Shan Y 25 Cross-layer techniques for adaptive video streaming over wireless networks EURASIP Journal on Applied Signal Processing [9] Tsai M-F, Shieh C-K, Ke C-H, Deng D-J 2 Su-packet forward error correction mechanism for video streaming over wireless networks Multimedia Tools and Applications 47() [] Park K, Wang W 2 QoS-sensitive transport of real-time MPEG video using adaptive redundancy control Computer Communications [] Wu H, Claypool M, Kinicki R 25 Adjusting Forward Error Correction with Temporal Scaling for TCP-Friendly Streaming MPEG ACM Transactions on Multimedia Computing Communications, and Applications (4) [2] Yuan Y, Cockurn B, Sikora T, Mand M 27 Efficient allocation of packet-level forward error correction in video streaming over the Internet Journal of Electronic Imaging 6(2) -2 [3] Shih C H, Shieh C K, Hwang W S 28 A Transparent oss Recovery Scheme Using Packet Redirection for Wireless Video Transmissions EURASIP Journal on Advances in Signal Processing Article ID [4] i A 27 RTP payload format for generic forward error correction RFC59 IETF [5] Plummer W W 989 TCP Checksum Function Design ACM SIGCOMM Computer Communication Review 9(2) 95- Authors Min Chen, 978-3, Hengyang Hunan, China. Current position, grades: associate professor. University studies: Central South University, China. Scientific interest: network technology, wireless sensor network. Ang i, , Yueyang Hunan, China. Current position, grades: associate professor. University studies: Central South University. Scientific interest: Network technology, Wireless sensor network. 59