Research On Complexity-Controllable Joint Source Channel Coding

With the rapid development in wireless communication, multimedia applications in wireless environment draw great attention in the third and forth generation mobile communication. While, the processing capabilities insufficiency of mobile device compared to PC and the microprocessor's power-constraint problem caused by battery become the major restrictions on the development of multimedia applications. Sponsored by "Joint Source and Channel Adaptive Coding Technology Based on Wireless Channel", subproject of National 863 project "Digital Audio, Video Encoding, Transporting, Testing and Applying Model"(No. 2002AA119010), and "Research on Joint Power-Rate-Distortion Model for Wireless Multimedia Communication"( No.20045006071-18), Chenguang Project of Youth S&T, this dissertation is focusing on the establishment of complexity-controllable joint source channel coding (JSCC) system with related techniques optimization and complexity model definition under the constraint of the microprocessor's power. Together with the power control policy, the work aims to adjust the source/channel encoding/decoding complexity adaptively to achieve the best-effort visual quality when the power level is fluctuant or limited. Firstly, a power controllable JSCC system is designed with the complexity prototype definition for each subsystem. Compared to the existing encoding system only considering the energy consumed by source coding, the proposed system takes the source coding, channel coding and error control into consideration to make a good balance between algorithm complexity and energy consumption of microprocessor. The JSCC system consists of three subsystems as: source encoding/decoding subsystem, joint channel protection subsystem and complexity control subsystem. The basic schemes of every subsystem are proved to be advanced. Based on the corresponding relationship between power and algorithm complexity, we clarified that the major way to control power consumption is controlling algorithm complexity. The time complexity of an algorithm is defined for complexity measurement to shield the model from realization variety. Through the statistical data retrieval for the ratio of technique time complexity to system time complexity, the complexity prototypes of source encoding/decoding, channel encoding/decoding and error control are established with major complexity-controlling parameter specifying. Furthermore, the strategy of JSCC power control and the method of complexity optimization are also discussed briefly. Secondly, a complexity and spatiality scalable video coding (CSSVC) system is proposed for the source encoding/decoding subsystem. Compared to the existing spatially scalable video coding system, the proposed system integrates the good compression performance of H.264 in low frequency domain with the attractive advantages of in-band prediction in wavelet domain to provide the good coding efficiency while maintaining the H.264 fully compatible scalable bitstream. Because of the obvious Gaussian distribution properties concluded by series of analysis for high frequency subband, the efficiency of the coding techniques selected for CSSVC prototype is discussed carefully. Several technical improvements in reference reconstruction and motion compensation, reordering scan, intra prediction, and motion estimation for high frequency decompositions endure our system with average 0.04dB gain in PSNR and better visual quality compared with the H.264 main profile in the full resolution. For each technical improvement, the complexity variation is analyzed and complexity scalable schemes are investigated for different processing capabilities. Through the theoretical analysis, statistical data retrieval and performance test, all of the technical improvements are proved to be efficient. Thirdly, a complexity-controllable hybrid error control system is constructed for joint channel protection subsystem. Compared to the existing approaches, the proposed system is designed for the unstable feedback channel. In order to integrate the advantages of hybrid ARQ (HARQ) I and HARQ II, a novel HARQ scheme called hybrid independent retransmission and self-decodable ARQ is proposed with parameters optimization in data retransmission time, data retransmission delay, response retransmission time, and response retransmission delay for ARQ protocols in transmitter and receiver. Using the more appropriate parameters for the wireless communication, the optimized HARQ system shows lower end-to-end delay and more stable performance by comparing to the general scheme. Moreover, the complexity control strategy is also specified for the joint channel protection subsystem. Finally, the time complexity models are investigated and established for the source encoding/decoding, channel encoding/decoding and joint channel protection system. Major technical components, such as motion estimation (ME), motion compensation (MC), intra prediction, transform and context-adaptive binary arithmetic coding (CABAC), are analyzed to figure out the proportion they occupied in the system complexity respectively.Based on the theoretical analysis and statistical data retrieval, the time complexity of proposed technique improvements, as reordering scan, intra prediction and motion estimation, is described measurably. The complexity control capabilities of INTRA mode, INTER mode, coding subbands, ratio of non-zero blocks and quantization parameter are further discussed and the relative encoding/decoding complexity variation is measured for different coding schemes. The experimental results have proved accuracy of the models. For the time complexity model of rate compatible punctured turbo code (RCPT), complexity influenced by interleave, recursive systematic code (RSC) encoding/decoding and de-interleave are analyzed detailedly. Model amendment is made for joint channel protection system and the JSCC power controlling strategy is designed taking complexity relationship among subsystems, microprocessor capability, fast complexity control and bitstream match into consideration. It provides a complete solution for the joint power control.