| In this dissertation, we investigate efficient adaptation techniques for supporting media streaming over a wireless channel. Using level-crossing analysis, we first devise a scheme for characterizing the wireless channel as a continuous-time Markov chain. Fluid analysis is then used to provide closed-form expressions for the packet loss and delay performance. We also analyze the wireless effective bandwidth that is needed to guarantee a certain level of loss or delay performance. We then consider application-level adaptation and present several schemes for joint optimization of the source rate and channel code. Our schemes, in general, account for the variability in the bit rate of compressed video, the channel dynamics, and the occupancies of the receiver and transmitter buffers. They attempt to maximize the encoder's source rate while reducing the likelihood of starvation at the decoder. One of these schemes is cycle based, whereby the source rate and channel code are adapted according to the cyclic nature (good/bad) of the channel. Another scheme uses dynamic frame playback, whereby the frame display rate is tuned according to the buffer occupancy. Simulations are used to demonstrate the performance advantages of the proposed adaptation techniques. |
