Video over TETRA Employing MPEG-4 Visual Coding Standard(Multimedia Systems)

IEICE TRANS. COMMUN., VOL.E87–B, NO.4 APRIL 2004 990 PAPER Video over TETRA Employing MPEG-4 Visual Coding Standard Yoong-Choon CHANG†a) and M. Salim BEG†† , Nonmembers SUMMARY Video transmission over Terrestrial Trunked Radio (TETRA) mobile channel employing MPEG-4 visual coding standard is proposed in this paper. Detail parameters of the proposed systems are discussed in this paper. Performance of the proposed systems was evaluated in Average Peak Signal to Noise Ratio (APSNR) versus Signal to Noise Ratio (SNR) and Bit Error Rate (BER). In particular, the video quality that can be achieved at different channel conditions and employing different combinations of MPEG-4 visual error resilient tools is presented in this paper. Results obtained show that higher video bitrate does not necessarily lead to higher video quality at the receiver as the received video quality depends on the bit error pattern or the number of error free video packets. key words: multimedia coding, MPEG-4, wireless video communications, TETRA 1. Introduction With the rapid growth and deployment of mobile and wireless communication systems during the last decade, the need for transmission of multimedia information over such links has become an important application requirement. Transmission of video is particularly challenging in view of the high data rate of raw video. Video compression techniques are therefore to be used to reduce the bandwidth requirements and enable the transmission of video information over bandlimited wireless channels. On the other hand, wireless and mobile channels are typically noisy and suffer from a number of channel degradations due to multipath reflections and fading. The effect of channel errors on compressed video bit stream is particularly severe, due to the manner in which the video is compressed and coded. Hence it is extremely critical to employ error resilient coding techniques in order to get robust transmission of coded video data, and is therefore the motivation behind the work presented in the paper. A number of techniques have been proposed to attempt to reduce the sensitivity of compressed video data to corruption [1]. Many of these techniques focussed on H.263 for low bit rate applications. H.263+ now has a wide variety of error resilience options [2]. The advent of MPEG-4 recently has introduced a wide range of new features [3]. Transmission of video over practical public mobile sysManuscript received May 9, 2003. Manuscript revised October 21, 2003. † The author is with the Faculty of Engineering, Multimedia University, Jalan Multimedia, 63100, Cyberjaya, Selangor, Malaysia. †† The author is with the Department of Electronics Engineering, Aligarh Muslim University, Aligarh, India. a) E-mail: ycchang@mmu.edu.my tems and protocols have been reported recently by some researchers. Transmission of compressed video over GSM has been proposed in 1996 [4] but due to the limited data traffic rate of GSM, which is up to a maximum of 9.6 kbit/s, the proposed system was only on transmitting still image over GSM channel. In 1999, video transmission over General Packet Radio Services (GPRS) was proposed [5]. In 2001, NTT Docomo, Japan has launched real-time visual and audio communications through third generation (3G) mobile networks. This paper however describes the proposed implementation of video transmission over a Terrestrial Trunked Radio (TETRA) which is a standard for Private Mobile Radio (PMR) [6], [7]. It may be noted that current TETRA phone only supports voice transmission function and TETRA phone with video transmission function is still under research and development to a large extent, and this is the motivation behind the work being reported here. A potential application for future TETRA phone with video transmission capability is the traffic accident incidence. A TETRA phone with integrated camera could be used to transmit live video from the traffic accident scene back to the police and ambulance services that could then accurately assess the situation and dispatch the relevant medical equipment and aid. The system being proposed by the authors in this paper has the following advantages which cannot be achieved with some of the previously proposed video over mobile systems: i) Group calls: The proposed system will be able to send video to other users in the same area. ii) Fast call set-up: The proposed system will be able to send video to other TETRA user almost immediately, without delay. This function is particularly useful and important during emergency situation. iii) Coverage: In the case where cellular service is disrupted during a nationwide power failure or due to some other reason, the proposed system will be very useful. Considering a bank raid case during such type of circumstances, police officers away from the scene can receive live video scene of the bank with the help of the police in the bank through the proposed video over TETRA system. Details of video transmission over TETRA employing MPEG-4 visual coding standard is proposed in this paper. Advantages of the proposed system compared with previous video over mobile systems have been presented briefly in the paper. In Sect. 2, an overview of TETRA and MPEG-4 is given. Section 3 presents the proposed systems, followed by details of experiment setup in Sect. 4. Results are presented CHANG and BEG: VIDEO OVER TETRA EMPLOYING MPEG-4 VISUAL CODING STANDARD 991 Table 1 in Sect. 5, followed by conclusions in Sect. 6. 2. Main TETRA parameters. Overview of TETRA and MPEG-4 Visual 2.1 TETRA TETRA was developed with support of European Telecommunications Standard Institute (ETSI) by the co-operation of manufacturers, users, operators primarily to bring the benefits of standardization to Private Mobile Communications [7]. TETRA actually includes a set of standards for voice and data applications and direct mode operation. Specifically, Voice + Data (V+D) is the standard for both speech (in circuit switched mode) and data (both in circuit and packet switched mode) communications [6]. Packet Data Optimised (PDO) is the one for optimised data transmission in packet switched mode. Direct Mode Operation (DMO) is a procedure by which mobile units may communicate directly by using frequencies outside the control of network, and operability is guaranteed in spite of network traffic load and radio coverage conditions. The definition of TETRA system is confined to layers 1 to 3 of the Open Systems interconnection (OSI) seven layers and the main parameters are shown in Table 1 [7]. The access technique used is Time Division Multiple Access (TDMA). One TDMA frame includes 4 time slots and lasts 56.67 ms. The channel bitrate is 36 kbit/s and the adopted modulation scheme is π/4-DQPSK (Differential Quaternary Phase Shift Keying). Three data traffic channels (TCH) are provided depending on the amount of inherent error protection. These are: i) TCH/7.2 offering unprotected data at 7.2 kbit/s net data rate. ii) TCH/4.8 offering low protected data at 4.8 kbit/s net data rate. iii) TCH/2.4 offering high protected data at 2.4 kbit/s net data rate. Higher net data rates up to 28.8 kbit/s, 19.2 kbit/s or 9.6 kbit/s may be provided by allocating up to four traffic physical channels to the same communication [7]. The data rates offered in circuit mode data are shown in Table 2. TETRA accommodates three different depths of interleaving (N = 1, 4 or 8) which may be applied to the data traffic channels TCH/4.8 and TCH/2.4. 2.2 MPEG-4 Visual MPEG-4 [8] is a ISO/IEC standard, developed by MPEG, the committee which also developed the Emmy Award winning standards of MPEG-1 and MPEG-2. While MPEG1 and MPEG-2 video aimed at devising coding tools for CD-ROM and digital television respectively, MPEG-4 video aims at providing tools and algorithm for efficient storage, transmission and manipulation of video data in multimedia environments. The fact that the MPEG-4 video coding scheme uses compression techniques makes any MPEG-4 based video Table 2 Summary of circuit mode data rates. communications application very vulnerable to errors or losses in transmitted video bitstream. In the absence of any error propagation control mechanism, the loss of each unit of information may cause the loss of information up to the next synchronisation point, e.g. Video Object Plan (VOP) headers. This phenomenon is known as spatial impairment propagation. Furthermore, due to the predictive nature of the MPEG-4 algorithm, when errors or losses occur in an I or P-Video Object Plane, the VOPs encoded using the affected I or P-VOP as reference will be decoded wrongly. The errors or losses will propagate until the next intra-coded VOP and this is known as temporal impairment propagation. These phenomena will affect the quality of the decoded MPEG-4 coded video at receiver. To address these issues, MPEG-4 video coding standard has defined a set of error resilient tools, namely Video Packet Resynchronisation Marker (RM), Data Partitioning (DP), Reversible Variable Length Codes (RVLC) and Header Extension Codes (HEC) [9]. MPEG-4 was chosen to be the video coding algorithm for this proposed system due to the following reasons: i) High coding efficiency: MPEG-4 was one of the video coding algorithms which was having a very high coding efficiency. ii) Error resilience: MPEG-4 has some error resilience functions which make it suitable to be used for video transmission over error-prone wireless channel. 3. Proposed Systems The proposed systems of video over TETRA employing IEICE TRANS. COMMUN., VOL.E87–B, NO.4 APRIL 2004 992 MPEG-4 visual coding standard is shown in Figs. 1 and 2 while the main systems parameters are shown in Table 3. As can be seen from Figs. 1 and 2, the proposed TETRA phone with built-in MPEG-4 visual simple profile codec encodes the input raw video to 14 kbit/s and 21 kbit/s and transmits the encoded video bitstream over three TCH/4.8 and TCH/7.2 channel time slots respectively. The proposed picture size is Quarter Common Intermediate Format (QCIF) of spatial resolution 176 × 144 pixels as this picture size is sufficient for simple video communications purpose. The encoded frame sequence proposed is IPPPIPPPPP. . . , which means the first and the fifth frames are Intra-coded frames (or I-frames) while the rest of the frames are Predictive-coded frames (or P-frames). 4. Fig. 1 Proposed system of video over TETRA employing MPEG-4 visual coding standard with the use of three TCH/4.8 channel time slots. Fig. 2 Proposed system of video over TETRA employing MPEG-4 visual coding standard with the use of three TCH/7.2 channel time slots. Table 3 Simulation Environment A software simulation of the proposed systems shown in Fig. 1 and Fig. 2 has been carried out in order to analyse its performance. Generally speaking, the proposed TETRA phone with video transmission capability is likely to be used for videoconferencing type of applications, and therefore in this work the two video sequences used are Sales and Grandma which are slow motion sequences. A snapshot of Sales and Grandma video sequences are shown in Figs. 3 and 4 respectively. For the case of video transmission over TETRA employing MPEG-4 visual coding standard with the use of three TCH/4.8 time slots, 14 kbit/s MPEG-4 encoded video was then channel coded through Rate Compatible Punctured Convolutional (RCPC) codes of rate 2/3, according to TETRA TCH/4.8 data traffic channel specification. It was then interleaved, re-ordered and modulated for transmission using TETRA π/4-DQPSK modulation, as shown in Fig. 5. For the case of video transmission over TETRA employing MPEG-4 visual coding standard with the use of three TCH/7.2 time slots, 21 kbit/s MPEG-4 encoded video was modulated using π/4-DQPSK modulation scheme and Fig. 3 A snapshot of sales video sequence. Proposed systems main parameters. Fig. 4 A snapshot of grandma video sequence. CHANG and BEG: VIDEO OVER TETRA EMPLOYING MPEG-4 VISUAL CODING STANDARD 993 Fig. 5 Basic components of the proposed video over TETRA employing MPEG-4 visual coding standard using three TCH/4.8 channel time slots. for a frame size of X ×Y pixels and T encoded video frames. MSE was then converted to peak signal-to-noise ratio (PSNR), in decibel (dB), where PSNR is defined as:   255 (2) PSNR = 20 log10 RMSE where root mean square error (RMSE) is the square root of the MSE of luminance component (Y) of the video sequence, 255 corresponds to peak-to-peak range of the encoded and decoded video signal (each quantised to 256 levels). PSNR increases with increasing picture quality. Results of twenty five simulations, performed with different multipath Rayleigh fading seeds, have been averaged in order to obtain more reliable results. The average PSNR is given by [10]: 1
PSNR(s) AVERAGE PSNR = 25 s=1 Fig. 6 Basic components of the proposed video over TETRA employing MPEG-4 visual coding standard using three TCH/7.2 channel time slots. transmitted over wireless channel, as shown in Fig. 6. The propagation conditions were those specified in TETRA as TU50 (Typical Urban at 50 km/hr) at 400 MHz carrier frequency. The TU channel model represents the multipath propagations found in typical urban conditions, and in this experiment, a mobile terminal velocity of 50 km/hr was assumed. At the receiver side, the receiver was subject to Additive White Gaussian Noise (AWGN) and hard decision Viterbi channel decoding was used. We coded the sequence in a single scalability layer and considered rectangular objects, coincident with frames. The approach taken to measure the distortion at the decoder after transmission was to calculate the mean square error (MSE) between the received video signal d[x, y, t] and the original video signal i[x, y, t], as described in the following equation: MS E = Y T X    1


  i x, y, t − d x, y, t 2 XYT x=1 y=1 t=1 (1) 25 5. (3) Experiment Results Figures 7 to 22 show the results obtained for the proposed video over TETRA employing MPEG-4 visual coding standard in Average PSNR versus BER. Average PSNR versus signal-to-noise ratio (SNR) results have been shown in [11]. The following characteristics can be observed from Figs. 7 to 14: i) N = 1 interleaving depth results in the highest average PSNR, except for those without using error resilient tools, followed by N = 4 and N = 8 interleaving depths. Generally, the average PSNR for N = 1 interleaving depth is 1 dB to 2 dB higher than N = 4 interleaving depth and 2 dB to 3 dB higher than N = 8 interleaving depth. ii) Generally, the average PSNR for TCH/4.8 channel with N = 1 interleaving depth is 5 dB higher than that of TCH/7.2 channel. iii) During channel condition of BER=1 × 10−4 , IEICE TRANS. COMMUN., VOL.E87–B, NO.4 APRIL 2004 994 Fig. 7 Average PSNR versus BER for grandma sequence with the use of (RM + DP + RVLC) error resilient tools. Fig. 10 Average PSNR versus BER for grandma sequence without using error resilient tools. Fig. 8 Average PSNR versus BER for grandma sequence with the use of (RM + DP) error resilient tools. Fig. 11 Average PSNR versus BER for sales sequence with the use of (RM + DP + RVLC) error resilient tools. Fig. 9 Average PSNR versus BER for grandma sequence with the use of RM error resilient tool. Fig. 12 Average PSNR versus BER for sales sequence with the use of (RM + DP) error resilient tools. TCH/7.2 channel slightly outperforms TCH/4.8 channel with the use of RM, DP and RVLC combination and RM and DP combination, as can be seen in Figs. 7 and 8. If there is no error resilient tool being used or only RM is being used, TCH/4.8 channel will outperform TCH/7.2 channel during channel condition of BER=1 × 10−4 , although the bitrate for TCH/7.2 is 50% higher than TCH/4.8. This shows the ineffectiveness of RM for TCH/7.2, even at very low BER CHANG and BEG: VIDEO OVER TETRA EMPLOYING MPEG-4 VISUAL CODING STANDARD 995 Fig. 13 Average PSNR versus BER for sales sequence with the use of RM error resilient tool. Fig. 16 Average PSNR versus BER for 14 kbit/s grandma sequence transmitted over three TCH/4.8 time slots at N = 4 interleaving depth. Fig. 14 Average PSNR versus BER for sales sequence without using error resilient tools. Fig. 17 Average PSNR versus BER for 14 kbit/s grandma sequence transmitted over three TCH/4.8 time slots at N = 8 interleaving depth. Fig. 15 Average PSNR versus BER for 14 kbit/s grandma sequence transmitted over three TCH/4.8 time slots at N = 1 interleaving depth. Fig. 18 Average PSNR versus BER for 21 kbit/s grandma sequence transmitted over three TCH/7.2 time slots. condition. This reason for (i) to (iii) above is due to the difference in bit error pattern, which is shown in Figs. 23 and 24. Figure 23 shows the bit error pattern comparison between N = 1 and N = 8 interleaving depth for TCH/4.8 channel. As can be seen from this figure, the duration of error free period for N = 1 is much longer than N = 8. During one error free period for N = 1 interleaving depth, there maybe IEICE TRANS. COMMUN., VOL.E87–B, NO.4 APRIL 2004 996 Fig. 19 Average PSNR versus BER for 14 kbit/s sales sequence transmitted over three TCH/4.8 time slots at N = 1 interleaving depth. Fig. 20 Average PSNR versus BER for 14 kbit/s sales sequence transmitted over three TCH/4.8 time slots at N = 4 interleaving depth. Fig. 21 Average PSNR versus BER for 14 kbit/s sales sequence transmitted over three TCH/4.8 time slots at N = 8 interleaving depth. two or three groups of error occurring for N = 8 interleaving depth. This means that the chance for error free video packet using N = 1 is higher than N = 8 and thus, higher video quality or average PSNR at the receiver. On the other hand, Fig. 22 Average PSNR versus BER for 21 kbit/s sales sequence transmitted over three TCH/7.2 time slots. Fig. 23 Bit error pattern comparison between N = 1 and N = 8 interleaving depth for TCH/4.8 channel. Fig. 24 Bit error pattern comparison between TCH/4.8 N = 1 interleaving depth and TCH/7.2 channel. N = 8 has more occurrence of errors and this leads to more video packets being corrupted in the process of transmitting through TETRA channels and consequently, a lower video quality or average PSNR than N = 1 interleaving depth. As shown in Fig. 24 for comparison between TCH/7.2 CHANG and BEG: VIDEO OVER TETRA EMPLOYING MPEG-4 VISUAL CODING STANDARD 997 and TCH/4.8 N = 1 interleaving depth, the earlier explanation applies here also. The burst error of TCH/4.8 N = 1 interleaving depth leads to longer error free periods and consequently, lesser video packets will be corrupted and thus, higher video quality or average PSNR at the receiver. This shows that a higher video bitrate does not necessarily lead to higher video quality at the receiver as the received video quality depends on the bit error pattern or the number of error free video packets. In other words, the higher the number of error free periods, the more uncorrupted video packet will be and this will lead to higher video quality at the receiver. Figures 15 to 22 show the average PSNR relation to BER for 14 kbit/s and 21 kbit/s MPEG-4 encoded Sales and Grandma sequence transmitted over Three TETRA TCH/4.8 and TCH/7.2 channel time slots by using different combinations of MPEG-4 video built-in error resilient tools. The following characteristics can be observed from these figures: i) Average PSNR obtained with the use of MPEG-4 video built in error resilient tools is much higher than those without MPEG-4 video built in error resilient tools. ii) Average PSNR for RM, DP and RVLC combination is only 0 dB to 1 dB higher than that obtained with RM and DP combination at the same BER condition. In other words, the use of RVLC will only increase the average PSNR slightly and this shows the ineffectiveness of RVLC. The ineffectiveness of RVLC may be due to the reason that RVLC is being used in texture part of the video packet and not much data can be recovered in the reverse direction when the texture part of the video packet is corrupted. iii) Average PSNR for RM and DP combination is significantly higher than that obtained by using RM only. Up to 8 dB gain in average PSNR can be achieved with RM and DP combination compared with the use of RM only. This shows the effectiveness of DP, which partitions the video packet into two parts, motion part and texture part. When the texture part of the video packet is corrupted, motion part data can still be used to decode the video bitstream but this will not be possible without DP. This is the reason why average PSNR obtained with DP is significantly higher than that without DP. iv) Average PSNR with RM is significantly higher than that without error resilient tools, in which up to 8 dB gain in average PSNR can be achieved with the use of RM. This is due to the reason that RM is effective at localising the error and regain synchronisation once an error is detected in the video bitstream. 6. Conclusions Video transmission over TETRA employing MPEG-4 visual coding standard is proposed in this paper. A software simulation work of the proposed systems has been carried out in order to analyse the performance of the proposed systems. Besides that, a comparative study of different combinations of MPEG-4 video built-in error resilient tools on TETRA mobile channel has been carried out and shown in this paper. Results obtained in this work have shown that higher video bit rate does not necessarily lead to higher video quality at the receiver because the received video quality depends on the bit error pattern and the number of error free video packets. Besides that, at most of the channel conditions, video quality that can be achieved with N = 1 interleaving depth is better than that achieved with N = 8 interleaving depth. These results will be extremely useful in the practical implementation of MPEG-4 video transmission over mobile channels, particularly transmission of slow motion video for video conferencing type of applications. Handling other video sequences with different properties and with very fast motion can be taken up for future work and further study. Acknowledgments The authors would like to thank Motorola Technology Sdn Bhd (Malaysia) for the funding on this project. References [1] Y. Wang and Q.-F. Zhu, “Error control and concealment for video communications: A review,” Proc. IEEE, vol.86, no.5, pp.974–997, 1998. [2] S. Wenger, G. Knorr, J. Ott, and F. Kossentini, “Error resilience support in H.263+,” IEEE Trans. Circuits Syst. Video Technol., vol.7, no.7, pp.867–877, 1998. [3] A. Puri and A. Elefthariadis, “MPEG-4: An object based multimedia coding standard supporting mobile applications,” ACM Mobile Networks Applications, vol.3, pp.5–32, June 1998. [4] J. Dunlop and E. Santos, “Transmission of compressed video over GSM data channels,” Proc. IEEE Veh. Technol. Conf., vol.1, pp.306–310, 1996. [5] S. Fabri and A. Cellatoglu, et al., “Transmission of multimedia services over GPRS using MPEG-4 codec video,” Proc. IEEE Veh. Technol. Conf., vol.1, pp.401–405, 1999. [6] ETSI EN 300 392-2, V.2.3.2 (2001-03), Terrestrial Trunked Radio (TETRA), Voice plus Data (V+D), Part 2, Air Interface. [7] J. Dunlop, S. Girma, and J. Irvine, Digital Mobile Communications and the TETRA System, John Wiley & Sons. [8] ISO/IEC 14496-2, Information Technology—Coding of Audio Visual Objects, Part 2: Visual, Dec. 1999. [9] R. Talluri, “Error-resilient video coding in the ISO MPEG-4 standard,” IEEE Commun. Mag., vol.36, no.6, pp.112–119, June 1998. [10] Y.C. Chang, M.S. Beg, and T.F. Tang, “Performance evaluation of MPEG-4 visual error resilient tools over a mobile channel,” IEEE Trans. Consum. Electron., vol.49, no.1, pp.7–13, Feb. 2003. [11] Y.C. Chang, “Performance evaluation of MPEG-4 coded video transmission over a TETRA mobile channel,” MEngSc (by research) Thesis, Multimedia University, April 2003. IEICE TRANS. COMMUN., VOL.E87–B, NO.4 APRIL 2004 998 Yoong Choon Chang obtained his B.Eng. (Hons) in Electrical & Electronic Engineering from University of Northumbria at Newcastle, UK in 1999 and MEngSc from Multimedia University, Malaysia in 2003. He was an Electrical Engineer in an Electrical & Mechanical contractor company at Malaysia from 1999 to 2001 before he joined Multimedia University, Malaysia in 2001 as a tutor. He is a lecturer at the Faculty of Engineering, Multimedia University, Malaysia since May 2003. M. Salim Beg obtained his B.Sc. Engg. (Electrical Engg.) from Aligarh University and M.E. from Indian Institute of Science, Bangalore. He was a Commonwealth Scholar at Loughborough University, England from where he got his Ph.D. in the area of Digital Communications over mobile links in 1990. His current research interest lies in the area of multimedia communications and signal processing; Wireless and mobile links; and Computer Networks. He has more than eighty publications in these areas. He has presented and chaired a number of sessions at International conferences in India and abroad. He is on the technical commitee of a number of international conferences. During last five years he was Associate Professor and Chairman, Centre for Multimedia Communications, Multimedia University, Malaysia. He has recently joined Aligarh Muslim University, Aligarh. View publication stats