Abstract
The aim of this thesis is to determine the robustness of certain variable length codes in the presence of channel errors. The codes studied are: Huffman, Huffman equivalent, and T-codes. The algorithm used to construct each code is presented, together with appropriate methods of modifying the constructed code. The synchronization properties of such codes are discussed.The robustness of each code is measured and compared by considering its average error recovery after an error has occured. Two theoretical models, based on Markov chains, are presented and contrasted and the most accurate one selected. A simulation of the codes has also been implemented, and observed error recovery averages recorded. Both the theoretical and practical observations are used to draw the necessary conclusions.
This work will demonstrate that Huffman equivalent codes are more robust than modified T-codes and Huffman codes. It will also be observed, that unmodified T-codes are competitive with Huffman equivalent codes. However, it is believed that unmodified T-codes are impractical as such codes are hard to construct or indeed nonexistent for many distributions.
Other observations will emphasize the accuracy of the chosen Markov model, and the resilience and stability of robust codes in a noisy channel. The theory will be shown to be consistent with the simulation and thus validated.
| Date of Award | Jan 2001 |
|---|---|
| Original language | English |
| Supervisor | Stephanie Perkins (Supervisor) & Derek Smith (Supervisor) |