Dr Ahmed Al-Ani

Image of Ahmed Al-Ani
Senior Lecturer, School of Elec, Mech and Mechatronic Systems
Core Member, Centre for Health Technologies
Associate Member, CIMS Research Strength
BSc (UT Baghdad), MPhil (QAU), PhD (QUT)
 
Phone
+61 2 9514 2420
Fax
+61 2 9514 2435
Room
CB02.06.04

Research Interests

Biomedical signal processing; and pattern classification

Can supervise: Yes

Book Chapters

Al-Ani, A. & Atiya, A. 2010, 'Pattern Classification using a Penalised Likelihood Method' in Schwenker, Friedhelm; El Gayar, Neamat (eds), Artificial Neural Networks in Pattern Recognition. Fourth IAPR TC3 Workshop., Springer, Germany, pp. 1-12.
View/Download from: UTSePress |
Penalized likelihood is a well-known theoretically justified approach that has recently attracted attention by the machine learning society. The objective function of the Penalized likelihood consists of the log likelihood of the data minus some term penalizing non-smooth solutions. Subsequently, maximizing this objective function would lead to some sort of trade-off between the faithfulness and the smoothness of the fit. There has been a lot of research to utilize penalized likelihood in regression, however, it is still to be thoroughly investigated in the pattern classification domain. We propose to use a penalty term based on the Knearest neighbors and an iterative approach to estimate the posterior probabilities. In addition, instead of fixing the value of K for all pattern, we developed a variable K approach, where the number of neighbors can vary from one sample to another. The chosen value of K for a given testing sample is influenced by the K values of its surrounding training samples as well as the most successful K value of all training samples. Comparison with a number of well-known classification methods proved the potential of the proposed method.
Khushaba, R.N., Al-Ani, A. & Al-Jumaily, A. 2010, 'Swarm based Fuzzy Discriminant Analysis for Multifunction Prosthesis Control' in Randy Goebel, J+rg Siekmann, WolfgangWahlster, Friedhelm Schwenker, Neamat El Gayar (eds), Lecture Notes in Artificial Intelligence 5998 - Artificial Neural Networks in Pattern Recognition, Springer, Germany, pp. 197-206.
View/Download from: UTSePress | Publisher's site
In order to interface the amputee++s with the real world, the myoelectric signal (MES) from human muscles is usually utilized within a pattern recognition scheme as an input to the controller of a prosthetic device. Since the MES is recorded using multi channels, the feature vector size can become very large. In order to reduce the computational cost and enhance the generalization capability of the classifier, a dimensionality reduction method is needed to identify an informative moderate size feature set. This paper proposes a new fuzzy version of the well known Fisher++s Linear Discriminant Analysis (LDA) feature projection technique. Furthermore, based on the fact that certain muscles might contribute more to the discrimination process, a novel feature weighting scheme is also presented by employing Particle Swarm Optimization (PSO) for the weights calculation. The new method, called PSOFLDA, is tested on real MES datasets and compared with other techniques to prove its superiority.
Khushaba, R.N., Al-Ani, A. & Al-Jumaily, A. 2009, 'Feature Subset Selection Using Differential Evolution' in Mario Koppen, Nikola Kasabov, and George Coghill (eds), Advances in Neuro-Information Processing - Lecture Notes in Computer Science, Springer, Germany, pp. 103-110.
View/Download from: UTSePress | Publisher's site
One of the fundamental motivations for feature selection is to overcome the curse of dimensionality. A novel feature selection algorithm is developed in this chapter based on a combination of Differential Evolution (DE) optimization technique and statistical feature distribution measures. The new algorithm, referred to as DEFS, utilizes the DE float number optimizer in a combinatorial optimization problem like feature selection. The proposed DEFS highly reduces the computational cost while at the same time proves to present a powerful performance. The DEFS is tested as a search procedure on different datasets with varying dimensionality. Practical results indicate the significance of the proposed DEFS in terms of solutions optimality and memory requirements.
Khushaba, R.N., Al-Ani, A., Al-Jumaily, A. & Nguyen, H.T. 2008, 'A Hybrid Nonlinear-Discriminant Analysis Feature Projection Technique' in Wobcke, Wayne; Zhang, Mengjie (eds), Lecture Notes In Computer Science Vol 5360: AI 2008 Advances in Artificial Intelligence, Springer, Germany, pp. 544-550.
View/Download from: UTSePress | Publisher's site
Feature set dimensionality reduction via Discriminant Analysis (DA) is one of the most sought after approaches in many applications. In this paper, a novel nonlinear DA technique is presented based on a hybrid of Artificial Neural Networks (ANN) and the Uncorrelated Linear Discriminant Analysis (ULDA). Although dimensionality reduction via ULDA can present a set of statistically uncorrelated features, but similar to the existing DA+s it assumes that the original data set is linearly separable, which is not the case with most real world problems. In order to overcome this problem, a one layer feed-forward ANN trained with a Differential Evolution (DE) optimization technique is combined with ULDA to implement a nonlinear feature projection technique. This combination acts as nonlinear discriminant analysis. The proposed approach is validated on a Brain Computer Interface (BCI) problem and compared with other techniques.

Conference Papers

Al-Ani, A. & Khushaba, R.N. 2012, 'A Population based Feature Subset Selection Algorithm Guided by Fuzzy Feature Dependency', The first International Conference on Advanced Machine Learning Technologies and Applications (AMLTA12), Cairo, Egypt, December 2012 in Advanced Machine Learning Technologies and Applications, ed Aboul Ella Hassanien, Abdel-Badeeh M. Salem, Rabie Ramadan, Tai-hoon Kim, Springer Berlin Heidelberg, Germany, pp. 430-438.
Population-based (or evolutionary) algorithms have been attracting an increased attention due to their powerful search capabilities. For the particular problem of feature selection, population-based methods aim to produce better +or fitter+ future generations that contain more informative subsets of features. It is well-known that feature subset selection is a very challenging optimization problem, especially when dealing with datasets that contain large number of features. Most of the commonly used population-based feature selection methods use operators that do not take into account relationships between features to generate future subsets, which can have an impact on their capabilities to properly explore the search space. We present here a new populationbased feature selection method that utilize dependency between features to guide the search. In addition, a novel method for estimating dependency between feature pairs is proposed based on the concept of fuzzy entropy. Results obtained from datasets with various sizes indicate the superiority of the proposed method in comparison to some of the wellknown methods in the literature.
Alsukker, A., Khushaba, R.N. & Al-Ani, A. 2011, 'Enhancing the diversity of genetic algorithm for improved feature selection', Istanbul, October 2010 in 2010 IEEE International Conference on Systems Man and Cybernetics (SMC), ed Erdal Kaynak, IEEE, USA, pp. 1325-1331.
View/Download from: UTSePress | Publisher's site
Genetic algorithm (GA) is one of the most widely used population-based evolutionary search algorithms. One of the challenging optimization problems in which GA has been extensively applied is feature selection. It aims at finding an optimal small size subset of features from the original large feature set. It has been found that the main limitation of the traditional GA-based feature selection is that it tends to get trapped in local minima, a problem known as premature convergence. A number of implementations are presented in the literature to overcome this problem based on fitness scaling, genetic operator modification, boosting genetic population diversity, etc. This paper presents a new modified genetic algorithm based on enhanced population diversity, parents' selection and improved genetic operators. Practical results indicate the significance of the proposed GA variant in comparison to many other algorithms from the literature on different datasets.
Alsukker, A.S. & Al-Ani, A. 2011, 'An Enhanced Neural Network Ensemble for Automatic Sleep Scoring', International Conference on Communications and Information Technology (ICCIT 2011), Aqaba, Jordan, March 2011 in International Conference on Communications and Information Technology (ICCIT 2011), ed Alsharaeh, E.; Kadri, A.; Karam, M., IEEE Xplore, Piscataway, USA, pp. 126-129.
View/Download from: UTSePress | Publisher's site
Improving the diversity of Neural Network Ensembles (NNE) plays an important role in creating robust classification systems in many fields. Several methods have been proposed in the literature to create such diversity using different sets of classifiers or using different portions of training/feature sets. Neural networks are often used as base classifiers in multiple classifier systems as they adapt easily to small changes in the training data, therefore creating diversity that is necessary to make the ensemble work. This paper presents a novel algorithm based on generating a set of classifiers such that each one of them is biased towards one of the target classes. This will improve the ensemble diversity and hence its performance. Results on sleep data sets show that the proposed method is able to outperform the traditional fusion algorithms of bagging and boosting.
Alsukker, A.S., Khushaba, R.N. & Al-Ani, A. 2010, 'Optimising the K-NN Metric Weights Using Differential Evolution', International Conference on Multimedia Computing and Information Technology, U.A.E., March 2010 in International Conference on Multimedia Computing and Information Technology (MCIT-2010), ed N/A, IEEE, USA, pp. 89-92.
View/Download from: UTSePress |
Traditional k-NN classifier poses many limitations including that it does not take into account each class distribution, importance of each feature, contribution ofeach neighbor, and the number ofinstances for each class. A Differential evolution (DE) optimization technique is utilized to enhance the performance of kNN through optimizing the metric weights offeatures, neighbors and classes. Several datasets are used to evaluate the performance of the proposed DE based metrics and to compare it to some k-NN variants from the literature. Practical experiments indicate that in most cases, incorporating DE in k-NN classification can provide more accurate performance.
Khushaba, R.N., Elliott, R., Alsukker, A., Al-Ani, A. & McKinley, S.M. 2010, 'Orthogonal locality sensitive fuzzy discriminant analysis in sleep-stage scoring', International Conference on Pattern Recognition, Istanbul, Turkey, August 2010 in Proceedings - 2010 International Conference on Pattern Recognition, ed NA, IEEE, United States, pp. 165-168.
View/Download from: UTSePress | Publisher's site
Sleep-stage scoring plays an important role in analyzing the sleep patterns of people. Studies have revealed that Intensive Care Unit (ICU) patients do not usually get enough quality sleep, and hence, analyzing their sleep patterns is of increased import
Alsukker, A.S., Al-Ani, A. & Atiya, A. 2009, 'A Modified K-Nearest Neighbor Classifier to Deal with Unbalanced Classes', International Conference on Neural Computation, Madeira, Portugal, October 2009 in International Conference on Neural Computation (ICNC 2009), ed Madani, K., INSTICC - Institute for Systems and Technologies of Information, Control and Communication, Portugal, pp. 408-413.
View/Download from: UTSePress |
We present in this paper a simple, yet valuable improvement to the traditional k-Nearest Neighbor (kNN) classifier. It aims at addressing the issue of unbalanced classes by maximizing the class-wise classification accuracy. The proposed classifier also gives the option of favoring a particular class through evaluating a small set of fuzzy rules. When tested on a number of UCI datasets, the proposed algorithm managed to achieve a uniformly good performance.
Khushaba, R.N., Al-Ani, A., Alsukker, A.S. & Al-Jumaily, A. 2008, 'A Combined Ant Colony and Differential Evolution Feature Selection Algorithm', International Workshop on Ant Colony, Brussels, Belgium, September 2008 in Lecture Notes In Computer Science Vol 5217: Ant Colony Optimization and Swarm Intelligence, ed Dorigo, M.; Birattari, M.; Blum, C.; Clerc, M.; St++tzle, Th.; Winfield, A., Springer, Belgium, pp. 1-12.
View/Download from: UTSePress | Publisher's site
Feature selection is an important step in many pattern recognition systems that aims to overcome the so-called curse of dimensionality problem. Although Ant Colony Optimization (ACO) proved to be a powerful technique in different optimization problems, but it still needs some improvements when applied to the feature selection problem. This is due to the fact that it builds its solutions sequentially, where in feature selection this behavior will most likely not lead to the optimal solution. In this paper, a novel feature selection algorithm based on a combination of ACO and a simple, yet powerful, Differential Evolution (DE) operator is presented. The proposed combination enhances both the exploration and exploitation capabilities of the search procedure. The new algorithm is tested on two biosignal-driven applications. The performance of the proposed algorithm is compared with other dimensionality reduction techniques to prove its superiority.
Khushaba, R.N., Al-Ani, A. & Al-Jumaily, A. 2008, 'Differential Evolution based Feature Subset Selection', International Conference on Pattern Recognition, USA, December 2008 in Proceedings of the 19th International Conference on Pattern Recognition (ICPR-2008), ed Anil K. Jain, IEEE, USA, pp. 1-4.
View/Download from: UTSePress | Publisher's site
In this paper, a novel feature selection algorithm based on differential evolution (DE) optimization technique is presented. The new algorithm, called DEFS, modifies the DE which is a real-valued optimizer, to suit the problem of feature selection. The proposed DEFS highly reduces the computational costs while at the same time proving to present powerful performance. The DEFS technique is applied to a brain-computer-interface (BCI) application and compared with other dimensionality reduction techniques. The practical results indicate the significance of the proposed algorithm in terms of solutions optimality, memory requirement, and computational cost.
Khushaba, R.N., Al-Ani, A. & Al-Jumaily, A. 2008, 'Fuzzy Discriminant Analysis based Feature Projection in Myoelectric Control', IEEE Engineering in Medicine and Biology Society Annual Conference, Canada, August 2008 in Proceedings of the 30th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, ed Dumont, G., IEEE, USA, pp. 5049-5052.
View/Download from: UTSePress | Publisher's site
The myoelectric signal (MES) from human muscles is usually utilized as an input to the controller of a multifunction prosthetic hand. In such a system, a pattern recognition approach is usually employed to discriminate between the MES from different classes. Since the MES is recorded using multi channels, the feature vector size can become very large. In order to reduce the computational cost and enhance the generalization capability of the classifier, a dimensionality reduction method is needed to identify an informative moderate size feature set. This paper proposes a novel feature projection technique based on a combination of Fisher's Linear Discriminant Analysis (LDA), and Fuzzy Logic. The new method, called FLDA, assigns different membership degrees to the data points thus reducing the effect of overlapping points in the discrimination process. Furthermore, the concept of Mutual Information (MI) is introduced in the fuzzy memberships in order to assign weights to the features (attributes) according to their contribution to the discrimination process. The FLDA method is tested on a seven classes MES dataset and compared with other feature projection techniques proving its superiority.
Alsukker, A.S., Khushaba, R.N., Al-Ani, A. & Al-Jumaily, A. 2008, 'Enhanced Feature Selection Algorithm Using Ant Colony Optimization and Fuzzy Memberships', IASTED International Conference on Biomedical Engineering, Innsbruck, Austria, February 2008 in Proceedings of the 6th IASTED International Conference on Biomedical Engineering (BioMED 2008), ed Hierlemann, A., IASTED, Canada, pp. 34-39.
View/Download from: UTSePress
Feature selection is an indispensable pre-processing step when mining huge datasets that can significantly improve the overall system performance. This paper presents a novel feature selection method that utilizes both the Ant Colony Optimization (ACO) and fuzzy memberships. The algorithm estimates the local importance of subsets of features, i.e., their pheromone intensities by utilizing fuzzy c-means (FCM) clustering technique. In order to prove the effectiveness of the proposed method, a comparison with another powerful ACO based feature selection algorithm that utilizes the Mutual Information (MI) concept is presented. The method is tested on two biosignals driven applications: Brain Computer Interface (BCI), and prosthetic devices control with myoelectric signals (MES). A linear discriminant analysis (LDA) classifier is used to measure the performance of the selected subsets in both applications. Practical experiments prove that the new algorithm can be as accurate as the original method with MI, but with a significant reduction in computational cost, especially when dealing with huge datasets.
Khushaba, R.N., Al-Ani, A. & Al-Jumaily, A. 2008, 'Swarm Intelligence In Myoelectric Control: Particle Swarm Based Dimensionality Reduction', IASTED International Conference on Biomedical Engineering, Austria, February 2008 in 6th IASTED International Conference on Biomedical Engineering (BioMED 2008), ed Hierlemann, A., IASTED, Canada, pp. 601-694.
View/Download from: UTSePress
The myoelectric signals (MES) from human muscles have been utilized in many applications such as prosthesis control. The identification of various MES temporal structures is used to control the movement of prosthetic devices by utilizing a pattern recognition approach. Recent advances in this field have shown that there are a number of factors limiting the clinical availability of such systems. Several control strategies have been proposed but deficiencies still exist with most of those strategies especially with the Dimensionality Reduction (DR) part. This paper proposes using Particle Swarm Optimization (PSO) algorithm with the concept of Mutual Information (MI) to produce a novel hybrid feature selection algorithm. The new algorithm, called PSOMIFS, is utilized as a DR tool in myoelectric control problems. The PSOMIFS will be compared with other techniques to prove the effectiveness of the proposed method. Accurate results are acquired using only a small subset of the original feature set producing a classification accuracy of 99% across a problem of ten classes based on tests done on six subjects MES datasets.
Khushaba, R.N., Alsukker, A.S., Al-Ani, A. & Al-Jumaily, A. 2008, 'Intelligent Artificial Ants based Feature Extraction from Wavelet Packet Coefficients for Biomedical Signal Classification', International Symposium on Communications, Control and Signal Processing, Malta, March 2008 in 3rd International Symposium on Communications, Control and Signal Processing (ISCCSP 2008), ed Maloberti, F; Micallef, P; Mitra, S.K., IEEE, USA, pp. 1366-1371.
View/Download from: UTSePress | Publisher's site
n this paper, a new feature extraction method utilizing ant colony optimization in the selection of wavelet packet transform (WPT) best basis is presented and adopted in classifying biomedical signals. The new algorithm, termed intelligent artificial ants (IAA), searches the wavelet packet tree for subsets of features that best interact together to produce high classification accuracies. While traversing the WPT tree, the IAA takes into account existing correlation between features thus avoiding information redundancy. The IAA method is a mixture of filter and wrapper approaches in feature subset selection. The pheromone that the ants lay down is updated by means of an estimation of the information contents of a single feature or feature subset. The significance of the subsets selected by the ants is measured using linear discriminant analysis (LDA) classifier. The IAA method is tested on one of the most important biosignal driven applications, which is the brain computer interface (BCI) problem with 56 EEG channels. Practical results indicate the significance of the proposed method achieving a maximum accuracy of 83%.
Darvishi, S. & Al-Ani, A. 2007, 'Brain-Computer Interface Analysis using Continuous Wavelet Transform and Adaptive Neuro-Fuzzy Classifier', IEEE Engineering in Medicine and Biology Society Annual Conference, Lyon, France, August 2007 in Proceedings of the 29th International Conference of the IEEE Engineering in Medicine and Biology Society, ed Andre Dittmar and John Clark, IEEE, Engineering in Medicine and Biology Society, USA, pp. 3220-3223.
View/Download from: UTSePress | Publisher's site
The purpose of this paper is to analyze the electroencephalogram (EEG) signals of imaginary left and right hand movements, an application of Brain-Computer Interface (BCI). We propose here to use an Adaptive Neuron-Fuzzy Inference System (ANFIS) as the classification algorithm. ANFIS has an advantage over many classification algorithms in that it provides a set of parameters and linguistic rules that can be useful in interpreting the relationship between extracted features. The continuous wavelet transform will be used to extract highly representative features from selected scales. The performance of ANFIS will be compared with the well-known support vector machine classifier.
Khushaba, R.N., Al-Ani, A. & Al-Jumaily, A. 2007, 'Swarm Intelligence based Dimensionality Reduction for Myoelectric Control', International Conference on Intelligent Sensors, Sensor Networks and Information Processing, Melbourne, Australia, December 2007 in The third International Conference on Intelligent Sensors, Sensor Networks and Information Processing (ISSNIP) 2007, ed M. Palaniswami, Slaven Marusic, and Yee Wei Law, IEEE, Melbourne, Australia, pp. 577-582.
View/Download from: UTSePress | Publisher's site
Recent approaches in multifunction myoelectrically controlled prosthetic devices revealed that dimensionality reduction plays a significant role in the overall system performance. In this paper, a new feature selection method is developed based on a mixture of particle swarm optimization (PSO) method and the concept of mutual information (MI). The new method, termed PSO-MI, is adopted as a dimensionality reduction tool for myoelectric control. The PSO-MI employs the MI measure to aid in controlling the movements of particles in the solution space, thus forming a kind of a hybrid filter-wrapper method. The new PSO-MI is able to account for the interaction property between the features in the selected subset, thus producing high classification accuracies. A dataset of transient myoelectric signal (MES) consisting of six classes of hand grasp is utilized to test the performance of the proposed method. It is proved that the PSO-MI outperforms other methods adopted for dimensionality reduction in myoelectric control achieving 95.5% of classification accuracy across six classes problem.
Khushaba, R.N., Al-Jumaily, A. & Al-Ani, A. 2007, 'Novel Feature Extraction Method based on Fuzzy Entropy and Wavelet Packet Transform for Myoelectric Control', International Symposium on Communications and Information Technologies, Sydney, Australia, October 2007 in Proceedings 7th International Symposium on Communications and Information Technologies (ISCIT '07), 2007, ed Eryk Dutkiewicz, IEEE, Sydney, Australia, pp. 352-357.
View/Download from: UTSePress | Publisher's site
In this paper, a novel feature extraction method based on the utilization of wavelet packet transform (WPT) and the concept of fuzzy entropy is presented. The method acts in steps, were in the first step the WPT is employed to generate a wavelet decomposition tree from which many features are extracted. In the second step, a new algorithm to compute the fuzzy entropy is developed and adopted as a measure of information content to judge on features suitability in classification, by setting a threshold and removing the features that fall under a certain threshold. In the third step, principle component analysis (PCA) is employed to reduce the dimensionality of the generated feature set. As an application, the new algorithm is employed in multifunction myoelectric control problem to prove its efficiency. Accurate results (99% accuracy) are acquired from using only a small subset of the original feature set generated by the wavelet tree. The obtained results indicate that the generated features are of maximum relevance and with minimum degree of redundancy.
Alsukker, A.S. & Al-Ani, A. 2006, 'Evaluation of Feature Selection Methods for Improved EEG Classification', International Conference on Biomedical and Pharmaceutical Engineering, Singapore, December 2006 in Proceedings of the International Conference on Biomedical and Pharmaceutical Engineering, ed Vadakke Matham Murukeshan, Research Publishing Services, Singapore, pp. 146-151.
View/Download from: UTSePress
Al-Ani, A. & Alsukker, A.S. 2006, 'Effect of Feature and channel Selection on EEG Classification', Annual International Conference of the IEEE Engineering in Medicine and Biology Society, New York City, USA, August 2006 in Proceedings of the 28th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, ed Andreas Hielscher, IEEE, New York City, USA, pp. 2171-2174.
View/Download from: UTSePress | Publisher's site
n this paper, we evaluate the significance of feature and channel selection on EEG classification. The selection process is performed by searching the feature/channel space using genetic algorithm, and evaluating the importance of subsets using a linear support vector machine classifier. Three approaches have been considered: (i) selecting a subset of features that will be used to represent a specified set of channels, (ii) selecting channels that are each represented by a specified set of features, and (iii) selecting individual features from different channels. When applied to a brain-computer interface (BCI) problem, results indicate that improvement in classification accuracy can be achieved by considering the correct combination of channels and features
Al-Ani, A. 2006, 'Channell and Feature Classifier Fusion in EEG Analysis', Cairo, Egypt, December 2006 in Proceedings of the 3rd Cairo International Biomedical Engineering Conference (CIBEC06), ed Mohamed Islam, Ahmed H. Kandil, Manal Abdel-Wahed, IEEE, Cairo Egypt, p. CV1.4.
Al-Ani, A. 2005, 'An Ant Colony Optimization Based Approach for Feature Selection', ICGT International Conference on Artifical Intelligence and Machine Learning, Cairo, Egypt, December 2005 in AIML 2005 Proceedings, ed Aboshosha; A.Dr rer.nat, The International Congress for Global Science and Technology (ICGST), Cairo, Egypt, pp. 1-6.
View/Download from: UTSePress | Publisher's site
This paper presents a new feature subset selection algorithm based on the Ant Colony Optimization (ACO). ACO is a metaheuristic inspired by the behaviour of real ants in their search for the shortest paths to food sources. It looks for optimal solutions by utilizing distributed computing, local heuristics and previous knowledge. We modified the ACO algorithm so that it can be used to search for the best subsets of features. A new pheromone trail update formula is presented, and the various parameters that lead to better convergence are tested. Results on speech classification problem show that the proposed algorithm achieves better performance than both greedy and genetic algorithm based feature selection methods.

Journal Articles

Al-Ani, A., Alsukker, A. & Khushaba, R.N. 2013, 'Feature subset selection using differential evolution and a wheel based search strategy', Swarm and Evolutionary Computation, vol. 9, no. 1, pp. 15-26.
View/Download from: UTSePress | Publisher's site
Differential evolution has started to attract a lot of attention as a powerful search method and has been successfully applied to a variety of applications including pattern recognition. One of the most important tasks in many pattern recognition systems is to find an informative subset of features that can effectively represent the underlying problem. Specifically, a large number of features can affect the system's classification accuracy and learning time. In order to overcome such problems, we propose a new feature selection method that utilizes differential evolution in a novel manner to identify relevant feature subsets. The proposed method aims to reduce the search space using a simple, yet powerful, procedure that involves distributing the features among a set of wheels. Two versions of the method are presented. In the first one, the desired feature subset size is predefined by the user, while in the second the user only needs to set an upper limit to the feature subset size. Experiments on a number of datasets with different sizes proved that the proposed method can achieve remarkably good results when compared with some of the well-known feature selection methods.
Gargiulo, G., Shephard, R., Tapson, J., Mcewan, A., Bifulco, P., Cesarelli, M., Jin, C., Al-Ani, A., Wang, N. & Van Schaik, A. 2012, 'Pregnancy Detection And Monitoring In Cattle Via Combined Foetus Electrocardiogram And Phonocardiogram Signal Processing', BMC Veterinary Research, vol. 8, no. NA, pp. 1-10.
View/Download from: UTSePress
Background: Pregnancy testing in cattle is commonly invasive requiring manual rectal palpation of the reproductive tract that presents risks to the operator and pregnancy. Alternative non-invasive tests have been developed but have not gained popularity
Al-Ani, A., Talaat, A.S., Atiya, A.F., Mokhtar, S.A. & Fayek, M. 2012, 'Multiclass Penalized Likelihood Pattern Classification Algorithm', Lecture Notes in Computer Science, vol. 7665, pp. 141-148.
View/Download from: UTSePress
Penalized likelihood is a general approach whereby an objective function is defined, consisting of the log likelihood of the data minus some term penalizing non-smooth solutions. Subsequently, this objective function is maximized, yielding a solution that achieves some sort of trade-off between the faithfulness and the smoothness of the fit. In this paper we extend the penalized likelihood classification that we proposed in earlier work to the multi class case. The algorithms are based on using a penalty term based on the K-nearest neighbors and the likelihood of the training patterns+ classifications. The algorithms are simple to implement, and result in a performance competitive with leading classifiers.
Al-Ani, A., Rabie, A., van Dun, B. & Dillon, H. 2012, 'Analysis of Alertness Status of Subjects Undergoing the Cortical Auditory Evoked Potential Hearing Test', Lecture Notes in Computer Science, vol. 7663, pp. 92-99.
View/Download from: UTSePress
In this paper, we analyze the EEG rhythms of subjects undergoing the cortical auditory evoked potential (CAEP) hearing test. Investigation of the importance of the different EEG rhythms in terms of their capability in differentiating between the different alertness states when considering 64 channel EEG montage is conducted. This is followed by considering subsets that contain 2, 3, 4 as well as all 5 EEG rhythms. Finally, a feature subset selection method based on differential evolution (DE) that has particularly been proposed to deal with multi-channel signals is used to search for the best subset of EEG rhythms for the various channels.
Ali, M.M., Al-Ani, A., Eamus, D. & Tan, D.K. 2012, 'A New Image-Processing-Based Technique for Measuring Leaf Dimensions', American-Eurasian J. Agric. & Environ. Sci., vol. 12, no. 12, pp. 1588-1594.
View/Download from: UTSePress
We propose in this paper a new method for measuring a number of leaf dimension parameters including height, width, average width, perimeter and area. A digital scanner is utilized to acquire leaf images, which unlike digital cameras, requires no calibration in terms of size and angle of the acquired images. Edge detection, filtering and thresholding algorithms are applied to identify the leaf section of the image against the background. Forty leaves that differ in shape and size were used to validate the estimated parameters against the true values and parameters produced by the popular Li-Cor 3100. Data indicated that the proposed method achieved a constantly high accuracy
Ali, M.M., Al-Ani, A., Eamus, D. & Tan, D.K. 2012, 'A New Image Processing Based Technique to Determine Chlorophyll in Plants', American-Eurasian J. Agric. & Environ. Sci., vol. 12, no. 10, pp. 1323-1328.
View/Download from: UTSePress
Leaf colour is usually used as a guide for assessments of nutrient status and plant health. We propose a new inexpensive, hand-held and easy-to-use technique for the detection of chlorophyll content and foliar nitrogen content in plants based on leaf colour. This method provides a rapid analysis and data storage at minimal cost and does not require any technical or laboratory skills. Most of the existing methods that examined relationships between chlorophyll status and leaf colour were developed for particular species. These methods acquire leaf images using digital cameras, which can be sensitive to lighting conditions (colour, angle, flux density) and hence, require proper calibration. Our method analyses leaf colour images obtained from a digital scanner that requires minimal calibration compared as it has its one light source and the angle and distance between light and leaf are constant. Our new algorithm produced superior correlations with the true value of foliar chlorophyll content measured in the laboratory compared with existing non-destructive methods when applied to three different species (lettuce, broccoli and tomato).
Khushaba, R.N., Al-Ani, A. & Al-Jumaily, A. 2011, 'Feature subset selection using differential evolution and a statistical repair mechanism', Expert Systems with Applications, vol. 38, no. 9, pp. 11515-11526.
View/Download from: UTSePress | Publisher's site
One of the fundamental motivations for feature selection is to overcome the curse of dimensionality problem. This paper presents a novel feature selection method utilizing a combination of differential evolution (DE) optimization method and a proposed repair mechanism based on feature distribution measures. The new method, abbreviated as DEFS, utilizes the DE float number optimizer in the combinatorial optimization problem of feature selection. In order to make the solutions generated by the float-optimizer suitable for feature selection, a roulette wheel structure is constructed and supplied with the probabilities of features distribution. These probabilities are constructed during iterations by identifying the features that contribute to the most promising solutions. The proposed DEFS is used to search for optimal subsets of features in datasets with varying dimensionality. It is then utilized to aid in the selection of Wavelet Packet Transform (WPT) best basis for classification problems, thus acting as a part of a feature extraction process. Practical results indicate the significance of the proposed method in comparison with other feature selection methods.
Khushaba, R.N., Al-Ani, A. & Al-Jumaily, A. 2011, 'Swarmed Discriminant Analysis for Multifunction Prosthesis Control', International Journal of Engineering and Natural Sciences, vol. 5, no. 1, pp. 27-34.
View/Download from: UTSePress
One of the approaches enabling people with amputated limbs to establish some sort of interface with the real world includes the utilization of the myoelectric signal (MES) from the remaining muscles of those limbs. The MES can be used as a control input to a multifunction prosthetic device. In this control scheme, known as the myoelectric control, a pattern recognition approach is usually utilized to discriminate between the MES signals that belong to different classes of the forearm movements. Since the MES is recorded using multiple channels, the feature vector size can become very large. In order to reduce the computational cost and enhance the generalization capability of the classifier, a dimensionality reduction method is needed to identify an informative yet moderate size feature set. This paper proposes a new fuzzy version of the well known Fisher's Linear Discriminant Analysis (LDA) feature projection technique. Furthermore, based on the fact that certain muscles might contribute more to the discrimination process, a novel feature weighting scheme is also presented by employing Particle Swarm Optimization (PSO) for estimating the weight of each feature. The new method, called PSOFLDA, is tested on real MES datasets and compared with other techniques to prove its superiority.
Khushaba, R.N., Al-Ani, A. & Al-Jumaily, A. 2010, 'Orthogonal Fuzzy Neighborhood Discriminant Analysis for Multifunction Myoelectric Hand Control', IEEE Transactions On Biomedical Engineering, vol. 57, no. 6, pp. 1410-1419.
View/Download from: UTSePress | Publisher's site
Developing accurate and powerful electromyogram (EMG) driven prostheses controllers that can provide the amputees with effective control on their artificial limbs, has been the focus of a great deal of research in the past few years. One of the major challenges in such research is extracting an informative subset of features that can best discriminate between the different forearm movements. In this paper, a new dimensionality reduction method, referred to as orthogonal fuzzy neighborhood discriminant analysis (OFNDA), is proposed as a response to such a challenge. Unlike existing attempts in fuzzy linear discriminant analysis, the objective of the proposed OFNDA is to minimize the distance between samples that belong to the same class and maximize the distance between the centers of different classes, while taking into account the contribution of the samples to the different classes. The proposed OFNDA is validated on EMG datasets collected from seven subjects performing a range of 5 to 10 classes of forearm movements. Practical results indicate the significance of OFNDA in comparison to many other feature projection methods (including locality preserving and uncorrelated variants of discriminant analysis) with accuracies ranging from 97.66% to 87.84% for 5 to 10 classes of movements, respectively, using only two EMG electrodes.
Khushaba, R.N., Al-Jumaily, A. & Al-Ani, A. 2009, 'Evolutionary Fuzzy Discriminant Analysis Feature Projection Technique in Myoelectric Control', Pattern Recognition Letters, vol. 30, no. 7, pp. 699-707.
View/Download from: UTSePress | Publisher's site
The controller of a multifunction prosthetic hand usually employs a pattern recognition scheme to discriminate between the myoelectric signals (MES) from different classes of the forearm movements. The MES is recorded using a multichannel approach that makes the feature vector size very large. Hence a dimensionality reduction technique is needed to identify an informative moderate size feature set. This paper proposes a novel feature projection technique based on a combination of fisher linear discriminant analysis (LDA), fuzzy logic (FL), and differential evolution (DE) optimization technique. The new technique, DEFLDA, assigns different membership degrees to the data points in order to reduce the effect of overlapping points in the discrimination process. Furthermore, an optimized weighting scheme is presented in which certain weights are assigned to the features according to their contribution in the discrimination process. The proposed DEFLDA is tested on different datasets and compared with other projection techniques to prove its functionality.
Atiya, A. & Al-Ani, A. 2009, 'A penalized likelihood based pattern classification algorithm', pattern recognition, vol. 42, no. 11, pp. 2684-2694.
View/Download from: UTSePress | Publisher's site
Penalized likelihood is a general approach whereby an objective function is defined, consisting of the log likelihood of the data minus some term penalizing non-smooth solutions. Subsequently, this objective function is maximized, yielding a solution that achieves some sort of trade-off between the faithfulness and the smoothness of the fit. Most work on that topic focused on the regression problem, and there has been little work on the classification problem. In this paper we propose a new classification method using the concept of penalized likelihood (for the two class case). By proposing a novel penalty term based on the K-nearest neighbors, simple analytical derivations have led to an algorithm that is proved to converge to the global optimum. Moreover, this algorithm is very simple to implement and converges typically in two or three iterations. We also introduced two variants of the method by distance-weighting the K-nearest neighbor contributions, and by tackling the unbalanced class patterns situation. We performed extensive experiments to compare the proposed method to several well-known classification methods. These simulations reveal that the proposed method achieves one of the top ranks in classification performance and with a fairly small computation time.
Khushaba, R.N., Alsukker, A.S., Al-Ani, A., Al-Jumaily, A. & Zomaya, A.Y. 2009, 'A Novel Swarm-Based Feature Selection Algorithm in Multifunction Myoelectric Control', Journal of Intelligent & Fuzzy Systems, vol. 20, no. 4/5, pp. 175-185.
View/Download from: UTSePress |
Accurate and computationally efficient myoelectric control strategies have been the focus of a great deal of research in recent years. Although many attempts exist in literature to develop such strategies, deficiencies still exist. One of the major challenges in myoelectric control is finding an optimal feature set that can best discriminate between classes. However, since the myoelectric signal is recorded using multi channels, the feature vector size can become very large. Hence a dimensionality reduction method is needed to identify an informative, yet small size feature set. This paper presents a new feature selection method based on modifying the Particle Swarm Optimization (PSO) algorithm with the inclusion of Mutual Information (MI) measure. The new method, called BPSOMI, is a mixture of filter and wrapper approaches of feature selection. In order to prove its efficiency, the proposed method is tested against other dimensionality reduction techniques proving powerful classification accuracy.
Al-Ani, A. 2009, 'A dependency-based search strategy for feature selection', Expert Systems with Applications, vol. 36, no. 10, pp. 12392-12398.
View/Download from: UTSePress | Publisher's site
Feature selection has become an increasingly important field of research. It aims at finding optimal feature subsets that can achieve better generalization on unseen data. However, this can be a very challenging task, especially when dealing with large feature sets. Hence, a search strategy is needed to explore a relatively small portion of the search space in order to find "semi-optimal" subsets. Many search strategies have been proposed in the literature, however most of them do not take into consideration relationships between features. Due to the fact that features usually have different degrees of dependency among each other, we propose in this paper a new search strategy that utilizes dependency between feature pairs to guide the search in the feature space. When compared to other well-known search strategies, the proposed method prevailed.
Khushaba, R.N., Al-Jumaily, A. & Al-Ani, A. 2009, 'Dimensionality Reduction with Neuro-Fuzzy Discriminant Analysis', International Journal of Computational Intelligence, vol. 5, no. 3, pp. 225-232.
View/Download from: UTSePress
One of the most important tasks in any pattern recognition system is to find an informative, yet small, subset of features with enhanced discriminatory power. In this paper, a new neuro-fuzzy discriminant analysis based feature projection technique is presented based on a two stages hybrid of Neural Networks, optimized with Differential Evolution (DE), and a proposed Fuzzy Linear Discriminant Analysis (FLDA) technique. Although dimensionality reduction via FLDA can present a set of well clustered features in the reduced space, but like any version of the existing DA+s it assumes that the original data set is linearly separable, which is not the case with many real world problems. In order to overcome this problem, the first stage of the proposed technique maps the initially extracted features in a nonlinear manner into a new domain, with larger dimensionality, in which the features are linearly separable. FLDA acts then on these linearly separable features to further reduce the dimensionality. The proposed combination, referred to as NFDA, is validated on a prosthetic device control problem with Electroencephalogram (EEG) datasets collected from 5 subjects achieving a maximum testing accuracy of 85.7% for a three classes of EEG based imaginations of movements.
Al-Ani, A. 2005, 'Feature Subset Selection Using Ant Colony Optimization', International Journal of Computational Intelligence, vol. 2, no. 1, pp. 53-58.
View/Download from: UTSePress
Al-Ani, A. & Deriche, M. 2004, 'Multi-Channel Subspace Mapping Using an Information Maximization Criterion', Multidimensional Systems and Signal Processing, vol. 15, no. 2, pp. 117-145.
View/Download from: UTSePress | Publisher's site
A new hybrid information maximization (HIM) algorithm is derived. This algorithm is able to perform subspace mapping of multi-channel signals, where the input (feature) vector for each of the channels is linearly transformed to an output vector. The algorithm is based on maximizing the mutual information (MI) between input and output sets for each of the channels, and between output sets across channels. Such formulation leads to a substantial redundancy reduction in the output sets, and the extraction of higher order features that exhibit coherence across time and/or space. In this paper, we develop the proposed algorithm and show that it combines efficiently the strengths of two well-known subspace mapping techniques, namely the principal component analysis (PCA) and the canonical correlation analysis (CCA). Unlike CCA, which is limited to two channels, the HIM algorithm can easily be extended to multiple channels. A number of simulations and real experiments are conducted to compare the performance of HIM to that of PCA and CCA.
Al-Ani, A., Deriche, M. & Chebil, J. 2003, 'A new mutual information based measure for feature selection', Intelligent Data Analysis, vol. 7, no. 1, pp. 43-57.
View/Download from: UTSePress
In this paper, we discuss the problem of feature selection and the importance of using mutual information in evaluating the discrimination ability of feature subsets between class labels. Because of the difficulties associated with estimating the exact value of mutual information, we propose a new evaluation measure that is based on the information gain and takes into consideration the interaction between features. The proposed measure is integrated into a robust feature selection scheme and compared with the well-known mutual information feature selection (MIFS) algorithm using the problems of texture classification, speech segment classification and speaker identification.