UC San Diego

We study the transmission of underwater video with better quality than the current literature. The underwater acoustic medium pathloss attenuation depends not only on transmission distance but also on the frequency occupied by the signal, where lower frequencies have lower attenuation for a given distance. We propose cross-layer design algorithms thatexploit this frequency-dependent attenuation by connecting channel reliability to the structure of the compressed video. The video data are categorized based on their importance. Orthogonal frequency division multiplexing is adopted as the modulation technique, such that different data can be sent on different frequencies. The underlying communications system accompanied with a noise analysis is developed in the thesis. We propose three new techniques and compare them to three baseline techniques. In the proposed techniques, important video information is transmitted on the least attenuated frequencies while less important data are transmitted through higher frequencies. Simulation results show that at least one of our proposed techniques can achieve significant improvements in the peak signal-to-noise ratio in comparison to the baseline techniques.

Then, we address the problem of imperfect resampling at the receiver, that can significantly degrade the system performance. While the conventional choice of Doppler compensation factor minimizes the intercarrier interference around the central subcarrier, we show that using a Doppler compensation factor that treats lower subcarriers preferentially in terms of intercarrier interference can significantly improve the system performance in the case of imperfect resampling.Two ad-hoc schemes are proposed in the thesis to find suitable Doppler compensation factors. In addition, different Doppler factors for the paths are considered. Our proposed schemes achieve higher peak-signal-to-noise ratio of the received video compared to conventional schemes.

Lastly, we propose a joint source-channel rate-distortion optimization for real-time video transmission. The encoder chooses the optimal quantization parameter, mode of each macroblock, and the optimal forward error correction code rate of each packet based on the channel state information at the beginning of the transmitted packet. Since the most common model for wireless channels is Rayleigh fading, the algorithm is tested for Rayleigh fading channels at different coherence time assumptions.