Iterative Joint Source-Channel Coding Techniques for Single and Multiterminal Sources in Communication Networks

In a communication system it results undoubtedly of great interest to compress the information generated by the data sources to its most elementary representation, so that the amount of power necessary for reliable communications can be reduced. It is often the case that the redundancy shown by a wide variety of information sources can be modelled by taking into account the probabilistic dependance among consecutive source symbols rather than the probabilistic distribution of a single symbol. These sources are commonly referred to as single or multiterminal sources “with memory” being the memory, in this latter case, the existing temporal correlation among the consecutive symbol vectors generated by the multiterminal source. It is well known that, when the source has memory, the average amount of information per source symbol is given by the entropy rate, which is lower than its entropy per single letter. In this context, given a coded or uncoded communication system, one can further decrease the power required to achieve a certain probability of error by taking into account this memory in the detection process, that is, by exploiting the entropy rate rather than the entropy per single letter of the correlated source. This Thesis focuses on this topic: concretely, we investigate the design of iterative encoding and decoding schemes for the transmission of single and multiterminal sources with memory through noisy point to point and Multiple Input Multiple Output (MIMO) channels. To that end, the dissertation is divided in two different (but closely related) parts: – The first part of the Thesis concentrates on the point to point transmission of single sources with memory. The classical way to tackle this problem is based on the Separation Theorem, i.e. by first implementing source compression and then channel coding. Assuming infinite complexity, no loss in performance is incurred when compared to joint source-channel coding techniques. However, when the complexity is an issue this separation is no longer optimal. In order to alleviate this lost in performance, joint source channel coding schemes have been proposed in the literature. Basically these schemes exploit the memory of the source by properly attaching the statistical structure of the source to the decoding process in the receiver and thus, the complexity of the decoding algorithm depends strongly on the characteristics of the source. In order to overcome this issue and relax the complexity of the decoder at the cost of increasing the complexity at the transmitter side, we show that the preprocessing of the source output prior to the encoding process by means of a data sorting algorithm yields an alternate universal method to exploit that correlation without relying on the parameters of the source at hand. Concretely, we investigate the application of the reversible Burrows Wheeler Transform (BWT) to both source (compression) and channel coding. For this latter case, we present a novel source controlled binary modulation scheme that adapts the allocated energy in accordance with the first order statistical distribution of the binary symbols at the output of the BWT. Simulation results show that the proposed scheme outperforms, for the particular evaluated cases, the aforementioned standard arrangements. – The second part of the Thesis places its attention on the transmission of multiterminal sources with memory through multiple user communication networks. We first deal with the gaussian broadcast channel, where each component of the output vector generated by the multiterminal source is sent to the corresponding receiver by using a single transmit signal. To that purpose, several encoding strategies are studied. The following main conclusion is drawn: the best performance is achieved by using a superposition encoding scheme, and using the existing memory of the multiterminal source in the decoding process of the information intended to the best receiver (i.e. the receiver in the broadcast channel with less input noise). Secondly, the exploitation of the memory in spatially correlated multiterminal sources when being sent through multiple access channels (MAC) is also covered, with special emphasis on the design of iterative receivers by means of factor graphs. We consider frequency selective MAC channels and propose joint source-channel coding schemes that combine equalization, decoding and exploitation of the memory in an iterative fashion based on assembling the corresponding factor graphs. Moreover, two different equalization criterions are considered, and an iterative estimation method for the parameters defining the correlation is easily integrated into the decoding process. Simulation results show that the performance of the derived schemes is close to the fundamental limits under the assumption that Shannon’s Separation Theorem holds. Summarizing the above contributions, the final goal of this dissertation lies on the design of signal processing communication schemes that take into account the temporal correlation (memory) of a single or multiterminal source so as to reduce the transmit power required for a certain level of performance and complexity.