Hybrid Digital-Analog Joint Source-Channel Coding

From analog television to digital television, wireless video transmission system has experienced a major technical innovation. Conventional digital wireless video trans-mission scheme usually consists of the separate digital video-compression part (such as quantization coding, entropy coding, etc.) and digital channel-coding part. Such a separate digital coding framework is designed based on Shannon’s source-channel separation theorem, and it is very suitable for the scenario of a single sender-receiver pair communicated over a static channel. But, for the time-varying channel, when the channel quality is lower than the target channel quality, due to the collapse of channel coding, separate coding is unable to decode any information; on the other hand, when the channel quality is higher than the target quality, due to adopting quantization coding, separate coding cannot further improve the reconstruction quality any more. This is the well-known cliff effect. Owing to the cliff effect, the separate coding is not suit for mo-bile video communication and video broadcast communication scenarios. Therefore, it is very necessary and significant to design joint source channel coding scheme for the mobile video communication and video broadcasting communication. In addition, on the theory of joint source channel coding, there are a lot of open problems, for ex-ample, for a broadcast communication scenario, given the average channel costs (e.g., transmitting power), what is the achievable distortion region of all the receivers?This dissertation focuses on the theoretical research of hybrid digital-analog (H-DA) based joint source channel coding, and its application in video communication. At first, for a memoryless source communicated over a memoryless point-to-point chan-nel, we derive the cost-distortion region achievable by HDA coding, and design the numerical algorithm to compute the optimal HDA coding and the corresponding per-formance. Second, we extend the results on point-to-point communication to broadcast communication and other network communication, in the same way, we also derive the cost-distortion region achievable by HDA coding for these cases. Next, we apply the theory of HDA coding derived previously to wireless video transmission scenarios. Finally, we apply HDA coding to cooperative video communication scenarios as well.The main innovations and contributions of this dissertation are listed as follows.1) On the theory of joint source channel coding, there exist a lot of open problems, and many basic problems remain to be solved, such as for a given any source-channel pair in a point-to-point communication system, besides separate coding, does there ex-ist any other optimal coding scheme that can achieve Shannon limit? In addition, for a given any source-channel pair in multi-user communication systems, what is the region of the achievable cost-distortion tuple? Solving these problems usually involves high dimensional source-channel coding, and moreover unlike separate source-channel cod-ing, these high dimensional code cannot be simplified to the calculation of single-letter quantities (i.e., entropy, mutual information, etc.), hence these problems are very diffi-cult to solve. In this dissertation, we study a special coding framework, HDA source-channel code. The reason why we consider the HDA coding here is that, on the one hand, HDA coding is simple to design and easy to analyze, and moreover it achieves the currently best performance; on the other hand, for many communication scenar-ios, it has been proved that HDA coding scheme is indeed theoretically optimal. In this dissertation, we consider point-to-point communication, broadcast communication, and other network communication systems, then give the achievable cost-distortion region by HDA coding, and the corresponding HDA coding scheme. From the results, it can be seen that the digital part of the optimal HDA coding scheme can be simplified to the calculation of single-letter quantities like separate source-channel coding, and the ana-log part can adopt a low dimensional mapping, hence the performance of HDA coding is computable. Furthermore, the relevant results can be applied to the practical commu-nication, such as video communication.2) In this dissertation, we apply the theoretical results on HDA coding to the practical wireless video communication, and then propose a hybrid digital-analog joint source-channel coding scheme, WSVC. Traditional video transmission schemes main-ly consist of analog coding schemes and digital coding schemes, but the HDA cod-ing scheme proposed in this dissertation is a novel and efficient video transmission scheme. This HDA scheme integrates the advantages of digital coding and analog cod-ing-high coding efficiency and continuous quality scalability (graceful quality varia-tion with channel quality varying over time); moreover, it is able to broadcast one video with different resolutions to fit the devices with different display resolutions. Relative to most of state-of-the-art video delivery methods, our WSVC has remarkable perfor-mance improvement in quality scalability, multicast efficiency and spatial and temporal scalability. Therefore, it is very suitable for wireless video broadcast/multicast and mo-bile video scenarios. In addition, usually, quality, temporal and spatial scalability are the concepts of video compression coding (or cable video communication), and we are the first to propose these scalability in the wireless video communication. As far as we know, our WSVC scheme is also the first scalable video coding scheme that realizes temporal and spatial scalability, as well as continuous quality scalability.3) In this dissertation, we apply the idea of HDA coding to the cooperative wireless video communication scenario, and first propose a hybrid digital-analog joint source-channel coding scheme, WCVC, for cooperative communication system. Similar to WSVC, the HDA coding integrates the advantages of digital coding and analog coding-high coding efficiency and high quality scalability. Compared to most of state-of-the-art video delivery methods, our WCVC has remarkable performance improvement in quality scalability and coding efficiency. Furthermore, it is compatible with WSVC, and consequently it can be easily extended to realize full scalability (quality, temporal and spatial scalability). Therefore, it is very suitable for cooperative wireless video broadcast/multicast and mobile video scenarios.