CROSS-LAYER FRAME SYNCHRONIZATION FOR H.264 VIDEO OVER WIMAX Maria G. Martini and Chaminda T. E. R. Hewage WMN, MINT Centre Faculty of CISM, Kingston University London Penrhyn Road, KT1 2EE, UK ABSTRACT In the WIMAX (802.16e) standard several small medium ac- cess control (MAC) packets may be aggregated to form a MAC burst, forwarded to the physical (PHY) layer [1] The burst transmission is a mechanism to allow multiple MAC protocol data units (PDUs) belonging to the same video chan- nel to be transmitted / received in an aggregated way, in order to enable power saving at the mobile station, by putting the transceiver in sleep mode during off-burst interval. As a con- sequence of it, at the receiver side individual packets should be isolated within a burst, i.e. frame synchronization should be performed. We propose here to forward soft values from the PHY to the MAC layer, and to use our previous results on frame synchronization based on soft values [2] [3]. In addi- tion, we exploit here a-priori information on the prevalence of 1’s or 0’s in the video bitstream and in packet headers, by performing application-aware and channel-aware MAC layer frame synchronization. Results show that the proposed ap- proach results in an evident gain, at the expense of a small additional complexity. Index Terms— Wimax, burst mode, cross-layer design, frame synchronization, soft values, a-priori information. 1. INTRODUCTION In the mobile WIMAX (802.16e) standard several small MAC packets may be aggregated to form a MAC burst, forwarded to the PHY layer [1]. The burst transmission is a mechanism to allow multiple MAC PDUs belonging to the same video chan- nel to be transmitted/ received in an aggregated way. Burst transmission enables mobile subscriber station (MSS) to save power by putting transceiver in sleep mode during off-burst interval. At the receiver side, individual packets should be isolated within a burst, i.e. frame synchronization has to be performed. The most widely used method for providing frame syn- chronization is to insert a fixed symbol pattern or sync word (SW) into the data stream. The receiver obtains frame syn- chronization by locating the position of the sync word in the This work was partially supported by the European Commission under FP7 project INFSO-ICT-214625 OPTIMIX. received data stream. A first, intuitive approach for locating such position consists of correlating, the received signal with the expected SW, looking for the position where this correla- tion is maximum in the case of fixed length window, or com- paring the correlation with a threshold. The problem of frame synchronization has been widely studied in the literature in the case of equiprobable data sym- bols: the performance evaluation for frame synchronization with periodically embedded sync words searched through cor- relation in binary symmetric channels (BSC) has been stud- ied in [4], where also synchronization sequences with good aperiodic autocorrelation properties have been identified. The problem of optimum frame synchronization has been afforded in [5] on additive white gaussian noise (AWGN): the optimal metric for AWGN channel has been identified for the consid- ered case of fixed length frames and equiprobable data sym- bols. In [6] a performance evaluation through simulation of these metrics has been presented. Synchronization for un- known frame lengths is studied in [2, 7, 8]. In many practical situations the assumption of uniform data distribution is not realistic. For instance in discrete co- sine transform (DCT) coded video there is always a linear relationship between the coding bit rate and the percentage of zeros among the quantized transform coefficients [9], which is directly related to the image content and is a measure of picture complexity. Furthermore, variable length codes are used and consecutive runs of 0’s and 1’s are possible. For instance, MPEG-2, H.263, and MPEG-4 are generally based on fixed tables of variable length codes (VLC). Context-based Adaptive Binary Arithmetic Coding (CABAC) is a normative part of the ITU-T, ISO/IEC standard H.264/AVC for video compression [10]. If in the bitstream 1’s or 0’s are prevail- ing, this information can be exploited in the detection met- ric [11] [8] [2] [3]. In this paper we propose to perform frame synchroniza- tion at WiMAX MAC layer through the insertion of SWs in the burst to separate individual packets. Although classically sync word detection would be performed through correlation, we propose here to forward soft values from the PHY layer to the MAC layer, in order to use the optimal metric derived in [2]. Furthermore, we study the statistical distribution of 1’s and 0’s in the H.264 stream in order to exploit it as a-priori