Cross layer optimization for video transmission over Multiple Access Wireless Systems.

摘要

In this dissertation, we propose cross layer frameworks for video transmission over Multiple Access Wireless Systems. We first study the problem of video transmission over wireless Generalized Multi Carrier Code Division Multiple Access (GMC-CDMA) systems. A scalable video source codec is used and a multi rate setup is assumed, i.e., each video user is allowed to occupy more than one GMC-CDMA channels. Furthermore, each of these channels can utilize a different number of sub carriers. We propose a cross-layer optimization method to select the source coding rate, channel coding rate, number of sub carriers per GMC-CDMA channel and transmission power per GMC-CDMA channel given a maximum transmission power for each video user and an available chip rate.;Next, a joint source-channel coding technique is proposed for H.264/AVC video transmission over MIMO-OFDM wireless links. A set of fast time-varying and frequency selective fading channels are considered. The proposed pilot assisted transmission scheme multiplexes known symbols with information bearing data to estimate the channel in the presence of Inter symbol Interference (ISI). The channel is assumed to be unchanged for the duration of one OFDM block and change independently from one OFDM block to the other. The number of pilot symbols for each OFDM block is held constant and is dispersed throughout the block for efficient channel estimation. The video transmission system uses a jointly optimal Space Frequency (SF) code design and pilot placement scheme. RCPC codes are used for channel protection. The problem formulation selects the set of source, channel coding rates along with the optimal pilot placement for a total bitrate constraint and compares the performance of the optimal power allocation case with the equal power distribution cases. At the transmitter, we have developed a method for the estimation of the video distortion at the receiver for given channel conditions.;In another work, we provide a cross layer framework for the transmission of robust packet based H.264/AVC video over DS-CDMA wireless channels with near-ML multiuser detection using linear filters and reliability based processing. A cross layer framework is proposed for both underloaded and overloaded systems. In overloaded CDMA systems, when the number of users is greater than the processing gain, sub-optimal detection becomes more difficult because it is more difficult to separate the users because of the high correlation among the signatures. Such a scenario is of great interest because there are time that the processing gain cannot be increased due to bandwidth constraints.;The near ML multiuser detection approaches the optimum detector as follows. A reliability value is obtained for each user bit received by using the output of a multiuser zero-forcing linear filter. The reliability values are then used to generate error sequences which provides bit decisions close to optimum ML detection. The H.264/AVC video data is packetized, forward error corrected using Reed Solomon codes, interleaved, spread, modulated and transported through the RTP/UDP/IP protocol stack. For a given bitrate budget, a high motion video sequence will require more bits for source coding and hence lesser bits are left for channel protection than a low motion video. We have incorporated different detection schemes for high and low motion video. Our content aware near ML multiuser detector scheme prefers the users with high motion by choosing a higher value of the distinct error pattern from the near ML framework to better detect the data than the low motion video users. This causes a reduction in the complexity of the overall system compared to ML detection for all users, while maintaining a close performance with the ML detector.;Finally, an error concealment scheme is proposed to conceal an entirely lost frame in a compressed video bitstream due to errors introduced during transmission. The proposed scheme targets low bit rate video transmission applications using H.264/AVC. The motion field of the lost frame is first reconstructed by copying the co-located motion vectors and reference indices from the last decoded reference frame. After the motion field estimation of the missing frame, motion compensation is performed to reconstruct the frame. This technique reuses existing modules of the video decoder and it does not incur extra complexity compared to decoding a normal frame. It has also been adopted as a non-normative decoder option to the JM reference software at the JVT meeting in Poznan, Poland in July 2005. (Abstract shortened by UMI.).