Victor Buttigieg

Constructions for Variable-Length Error-Correcting Codes

Two construction techniques for variable-length error-correcting (VLEC) codes are given. The first uses fixed-length linear codes and anticodes to build new VLEC codes, whereas the second uses a heuristic algorithm to perform a computer search for good VLEC codes. VLEC codes may be used for combined source and channel coding. It is shown that over an additive white Gaussian noise channel the codes so constructed can perform better than standard cascaded source and channel codes with similar parameters.

Codebook and marker sequence design for synchronization-correcting codes

We propose a construction based on synchronization and error-correcting block codes and a matched marker sequence. The block codes can correct insertion, deletion and substitution errors within each codeword. The marker sequence allows the decoder to maintain synchronization at codeword boundaries even at high error rates. An upper bound is given for the performance of these codes over a channel with random substitutions and synchronization errors. It is shown that the performance is largely dependent on the codes minimum Levenshtein distance. The performance of these codes is verified by simulation and compared to published results. In concatenation with a non-binary outer code we obtain a significant improvement in frame error rate at similar overall code rates.

Maximum a Posteriori Decoding of Arithmetic Codes in Joint Source-Channel Coding

Arithmetic codes are being increasingly used in the entropy coding stage in many multimedia transmission applications. Combining channel coding with arithmetic coding can give implementation and performance advantages compared to separate source and channel coding. In this work, novel improvements are introduced into a technique by Grangetto et al. that uses maximum a posteriori (MAP) estimation for decoding joint source-channel coding using arithmetic codes. The arithmetic decoder is modified for quicker symbol decoding and error detection by the introduction of a look-ahead technique, and the calculation of the MAP metric is modified for faster error detection. These modifications also result in improved performance compared to the original scheme. Experimental results show an improvement of up to 0.4 dB when using soft-decision decoding and 0.6 dB when using hard-decision decoding.

Time-varying block codes for synchronisation errors: maximum a posteriori decoder and practical issues

In this study, the authors consider time-varying block (TVB) codes, which generalise a number of previous synchronisation error-correcting codes. They also consider various practical issues related to maximum a posteriori (MAP) decoding of these codes. Specifically, they give an expression for the expected distribution of drift between transmitter and receiver because of synchronisation errors. They determine an appropriate choice for state space limits based on the drift probability distribution. In turn, they obtain an expression for the decoder complexity under given channel conditions in terms of the state space limits used. For a given state space, they also give a number of optimisations that reduce the algorithm complexity with no further loss of decoder performance. They also show how the MAP decoder can be used in the absence of known frame boundaries, and demonstrate that an appropriate choice of decoder parameters allows the decoder to approach the performance when frame boundaries are known, at the expense of some increase in complexity. Finally, they express some existing constructions as TVB codes, comparing performance with published results and showing that improved performance is possible by taking advantage of the flexibility of TVB codes.

Interleavers for unpunctured symmetric turbo codes

The turbo code interleaver design problem is considered for relatively large block sizes, where the effect of trellis termination is less marked. An optimised interleaver design technique based on simulated annealing is proposed. The performance is significantly better than the Berrou-Glavieux interleaver without an increase in delay.

Using variable-length codes to correct Insertion, Deletion and substitution errors

A maximum likelihood metric is derived for the decoding of variable-length codes over a Binary Substitution, Insertion and Deletion channel. Using this metric a near-maximum likelihood decoder is derived. It is shown that variable-length codes can be used effectively to correct for insertion, deletion and substitution errors.

Design of an electric vehicle for the Maltese islands

This paper presents the design of an electric vehicle (EV) for operation over short distances (approximately 80 km) with an average speed of 80 km/h. These specifications were chosen to fulfil the daily transport needs of a small island such as Malta.

Spam Detection Using Linear Genetic Programming

Spam refers to unsolicited bulk email. Many algorithms have been applied to the spam detection problem and many programs have been developed. The problem is an adversarial one and an ongoing fight against spammers. We prove that reliable Spam detection is an NP-complete problem, by mapping email spams to metamorphic viruses and applying Spinellis’s [30] proof of NP-completeness of metamorphic viruses. Using a number of features extracted from the SpamAssassin Data set, a linear genetic programming (LGP) system called Gagenes LGP (or GLGP) has been implemented. The system has been shown to give 99.83% accuracy, higher than Awad et al.’s [3] result with the Naive Bayes algorithm. GLGP’s recall and precision are higher than Awad et al.’s, and GLGP’s Accuracy is also higher than the reported results by Lai and Tsai [19].

Construction of variable-length error-correcting codes using MOGA

Variable-length codes are classically used for source coding to compress data. However, Variable-Length Error-Correcting (VLEC) codes can also be used for channel coding where they are able to overcome the effects of channel impairments by detecting and correcting bit errors. Multi-objective optimization is an area of multiple criteria decision making, where an optimization algorithm attempts to solve a problem with more than one objective. In this paper, VLEC codes are constructed by taking into consideration their two main properties: Average Codeword Length and Free Distance, associated respectively with their source and channel coding aspects. These two properties are taken as objectives in a Multi-objective global optimization genetic algorithm. It is shown that better codes can be constructed using this algorithm.

A MAP Decoder for TVB Codes on a Generalized Iyengar–Siegel–Wolf BPMR Markov Channel Model

We present a generalization of the Iyengar-Siegel-Wolf Markov channel model for bit-patterned media recording media to allow negative drift, and adapt the maximum a posteriori decoder for time-varying block codes to work on this generalized model while minimizing complexity. We also describe a method for designing near-optimal codes for this channel using simulated annealing, obtaining better performance than alternative designs. In concatenation with a (999, 888)16 low-density parity-check (LDPC) code, we achieve a frame error rate (FER) of 10-3 at a channel error rate that is 1.73× higher than the best result with existing designs. A simple extension to include substitution errors allows the channel to approximate the dependent insertion, deletion, and substitution (DIDS) channel, with a decoding complexity that is 10× lower than that of Wu and Armands RC2 decoder. The performance in the absence of burst errors is almost identical. When the DIDS channel includes burst substitution errors, our decoder performs worse than the RC2 decoder, but maintains its complexity advantage. For the same concatenated code, our decoder achieves an FER of 10-3 at a channel error rate that is 1.68× lower than the RC2 decoder. Finally, simulation results show that our code designs improve on existing constructions for the DIDS channel.