Elizabeth C. Botha

Fingerprint recognition in low quality images

Abstract Algorithms are identified which are best suited for an automatic fingerprint recognition system operating on low quality images. New preprocessing algorithms for noise removal and binarization are described. Three approaches to classification are investigated: a correlation classifier, and two feature-based classification schemes. The best results on a database of 80 fingerprints are obtained with spatial-frequency features. Three classifiers (neural net, linear classifier and nearest neighbour) using these features are successful in identifying an independent test set. Details of the results are shown. In conclusion suggestions are made concerning the most suitable algorithms in each of the processing steps.

Cross-language use of acoustic information for automatic speech recognition

Techniques are investigated that use acoustic information from existing source language databases to implement automatic speech recognition (ASR) systems for new target languages. The assumption is that the amount of target language data available is too little for the training of a robust ASR system. Strategies for cross-language use of acoustic information are evaluated which include (i) training on pooled source and target language data, (ii) adapting source language models using target language data, (iii) adapting models trained on pooled source and target language using target language data only and (iv) transforming source language data to augment target language data for model training. These strategies are allied with Bayesian and transformation-based techniques to present a framework for cross-language reuse of acoustic information. Experiments are performed for a large number of approaches from the framework, using relatively large amounts of English speech data from either a separate database or from the same-database as smaller amounts of Afrikaans speech data to improve the performance of an Afrikaans speech recogniser. Results indicate that a significant reduction in word error rate is achievable (between 14% and 48% for experiments), depending on the amount of target language data available.

Related approaches to gradient-based thresholding

Abstract A recent segmentation method that uses transition regions to select an appropriate threshold for binarization is compared to an older averaged gradient thresholding method. It is shown that these methods are closely related to each other. We also show that this relation enables us to relieve some of the problems concerning edge-based thresholding.

Leap-frog is a robust algorithm for training neural networks

Optimization of perceptron neural network classifiers requires an optimization algorithm that is robust. In general, the best network is selected after a number of optimization trials. An effective optimization algorithm generates good weight-vector solutions in a few optimization trial runs owing to its inherent ability to escape local minima, where a less effective algorithm requires a larger number of trial runs. Repetitive training and testing is a tedious process, so that an effective algorithm is desirable to reduce training time and increase the quality of the set of available weight-vector solutions. We present leap-frog as a robust optimization algorithm for training neural networks. In this paper the dynamic principles of leap-frog are described together with experiments to show the ability of leap-frog to generate reliable weight-vector solutions. Performance histograms are used to compare leap-frog with a variable-metric method, a conjugate-gradient method with modified restarts, and a constrained-momentum-based algorithm. Results indicate that leap-frog performs better in terms of classification error than the remaining three algorithms on two distinctly different test problems.

Feature-based classification of aerospace radar targets using neural networks

The recognition of multi-frequency radar backscatter of non-cooperative aerospace targets, using feedforward neural networks, is investigated. Experiments are conducted on the complex backscatter of three model aircraft, measured in a compact range. For recognition, both down-range profiles (1-D) and ISAR (2-D) images of the targets are considered and features for both these representations are introduced. We propose a number of classification paradigms: a single feature-based classification, as well as more robust schemes. The results presented show that significant improvements in classification performance can be obtained through the integration of feature spaces and the combination of the outcomes of several classifiers.

The parallel downhill simplex algorithm for unconstrained optimisation

In this paper we present a parallel implementation of a well-known heuristic optimisation algorithm (the downhill simplex algorithm developed by Nelder and Mead in 1965) which is well suited for unconstrained optimisation. We present the sequential algorithm as well as the parallel algorithm which we used to generate numerical results. They include numerical results of experiments on neural networks and a test suite of functions which demonstrate the parallel algorithms increased robustness and convergence rate for high-dimensional problems compared to the sequential algorithm.

Automatic face recognition in a heterogeneous population

Abstract Most face recognition results in the open literature have used Caucasian faces to test their algorithms, while no special mention is made of the other population groups. The effect of a heterogeneous population on face recognition is therefore something of an unknown quantity. It is the aim of this paper to investigate the performance of a face recognition system, with emphasis on the effect of a heterogeneous population. We have found that there is no significant difference in the performance of the system on the different population groups.

Improved generalization of MCE parameter estimation with application to speech recognition

Discriminative training of hidden Markov models (HMMs) using minimum classification error training (MCE) has been shown to work well for certain speech recognition applications. MCE is, however, somewhat prone to overspecialization. This study investigates various techniques which improve performance and generalization of the MCE algorithm. Improvements of up to 10% in relative error rate on the test set are achieved for the TIMIT dataset.

Relative performance of correlation-based and feature-based classifiers of aircraft using radar range profiles

A surveillance system which automatically classifies aircraft can be very useful, especially with the problem of congestion at large airports. This paper evaluates the performance of correlation- and feature-based classifiers on a set of four simulated radar targets over a wide range of target orientations. Experiments are performed for a range of radar bandwidths in order to determine the effect of radar bandwidth on the relative classification performance. Only 1D radar range profiles are considered since it is assumed that the aircraft are classified using few profiles and the orientation of the aircraft in each profile is known only approximately. The results suggest that feature-based classifiers outperform correlation-based classifiers and that classification performance is highly dependent on the orientation of the aircraft, but that accurate classification of approaching aircraft is possible.

New rotation-invariant features for radar target recognition

Abstract Sophisticated modern radar systems have made it possible to build up detailed electromagnetic images of an object in the radar beam. This has generated interest in pattern recognition methods for radar target recognition. A feature of these high-resolution images is that some peaks relate to interactions between structures, as opposed to direct specular scattering. We propose a new method using interaction (multiple bounce) terms to compute a truly rotation-invariant feature for radar targets to be used for recognition. The effectiveness of the method is demonstrated by applying it to real data obtained from measurements in a compact range and demonstrating that the resulting feature sets could be used to effectively classify two similar target configurations.