Abstract
Video data claims a significant portion of global mobile data communications, currently standing at 55%. This demand outburst has been due to exceptional display technologies, on-demand video, gaming and live video streaming, to name a few. Despite the massive data rates supported by modern mobile communication technologies, video data is starting to overload mobile networks. This is particularly true in links with low connectivity, where repeat requests flood the system. As solutions for this inevitable demand growth, in addition to efficient video compression methods, more video data error resilience must be sought. One reason video traffic is vulnerable to channel errors is the method it is treated at transmission; treatment as any other generic data type. Video is a unique data type because its ultimate user is not a machine but a human, and the contents within the data are interdependent on each other. Based on its properties, video compression, transmission methodology, and the decoding function must be adapted. By considering video communication as a collaborative effort of these three functions, error resilience can be effectively implemented. Analysis of radio resources available for data transmission in a multipath fading channel reveals that some resources are more robust than others. In the first contribution of this thesis, this characteristic is utilised to impose more resilience to more sensitive data within the video. Reliable means of forecasting the relative robustness of each radio resource are designed. Criteria for identifying the sensitivity of different video data segments are formulated. Finally, a technology is presented to map data to radio resources such that maximum received video quality is achieved. While the focus of the first contribution was on harmonising the transmission methods with the features of the compressed video payload, the second contribution takes an alternative route to error resilience by focusing on the decoder. The compressed video payload entails some identifiable syntax elements, some of which follow a predictable pattern. This feature is exploited to improve error recovery at an iterative turbo decoder. An algorithm to identify the video frame boundaries in corrupted compressed sequences is formulated, along with algorithms to deduce the correct values for selected fields in the compressed stream. Modifying the turbo extrinsic information using these corrections act as reinforcements in the turbo decoding iterative process. Most communication protocols transmit data as blocks in an ordered sequence and await the acknowledgement of the receiver to determine the next block to be transmitted. This gives rise to latency issues and the overloading of the network when the link connectivity is poor. A solution is presented for video data in the final contribution, where the concept of a digital fountain is hired. A two-dimensional forward error correction strategy is introduced for a digital fountain, where first, the video payload is LDPC encoded and then turbo encoded. A joint decoding strategy is designed between the turbo decoder and the LDPC decoder to recover the video data in an iterative manner. Taken together, these contributions are solutions for the video data burden on mobile networks; solutions which reduce the necessity for re-transmissions. The presented error resilience techniques are updates to the existing transmission methodology and the decoding function. They explore a new paradigm of improving coverage and channel throughput.