Showing papers by "Aly A. Fahmy published in 2011"

Using advanced Hidden Markov Models for online Arabic handwriting recognition

Abstract: Online handwriting recognition of Arabic script is a difficult problem since it is naturally both cursive and unconstrained. The analysis of Arabic script is further complicated due to obligatory dots/stokes that are placed above or below most letters and usually are written delayed in order. This paper introduces a Hidden Markov Model (HMM) based system to provide solutions for most of the difficulties inherent in recognizing Arabic script. A preprocessing for the delayed strokes to match the structure of the HMM model is introduced. The used HMM models are trained with Writer Adaptive Training (WAT) to minimize the variance between writers in the training data. Also the models discrimination power is enhanced with Discriminative training. The system performance is evaluated using an international test set from the ADAB completion and shows a promising performance compared with the state-of-art systems.

Interval-based attribute evaluation algorithm

Abstract: Attribute values may be either discrete or continuous. Attribute selection methods for continuous attributes had to be preceded by a discretization method to act properly. The resulted accuracy or correctness has a great dependance on the discretization method. However, this paper proposes an attribute selection and ranking method without introducing such technique. The proposed algorithm depends on a hypothesis that the decrease of the overlapped interval of values for every class label indicates the increase of the importance of such attribute. Such hypothesis were proved by comparing the results of the proposed algorithm to other attribute selection algorithms. The comparison between different attribute selection algorithms is based on the characteristics of relevant and irrelevant attributes and their effect on the classification performance. The results shows that the proposed attribute selection algorithm leads to a better classification performance than other methods. The test is applied on medical data sets that represent a real life continuous data sets.

Feature evaluation based Fuzzy C-Mean classification

Abstract: Fuzzy C-Means Clustering, FCM, is an iterative algorithm whose aim is to find the center or centroid of data clusters that minimize an assigned dissimilarity function. The degree of being in a certain cluster can be defined in terms of the distance to the cluster-centroid. The domain knowledge is used to formulate an appropriate measure. However the Euclidean distance is considered as a general measure for such value. The calculation of the Euclidean distance doesn't take into consideration the degree of relevance of each feature to the classification model. In this paper, scoring methods like ChiMerge and Mutual information are used in the FCM model to improve the calculation of the Euclidean distance. Experimental results demonstrate the better performances of the improved FCM on UCI benchmark data sets rather than the ordinary FCM, where the ordinary FCM uses in classification either all features or the most important features while the improved FCM uses all the features but the Euclidean Distance will be calculated according to the relevance degree of each feature.

Heart Sound Feature Reduction Approach for Improving the Heart Valve Diseases Identification

Abstract: Recently, heart sound signals have been used in the detection of the heart valve status and the identification of the heart valve disease. Due to these characteristics, therefore, two feature reduction techniques have been proposed prior applying data classifications in this paper. The first technique is the chi-Square which measures the lack of independence between each heart sound feature and the target class, while the second technique is the deep believe network that uses to generate a new data set of a reduced number of features according the partition of the heart signals. The importance of feature reduction prior applying data classification is not only to improve the classification accuracy and to enhance the training and testing performance, but also it is important to detect which of the stages of heart sound is important for the detection of sick people among normal set of people, and which period important for the classification of heart murmur. Different classification algorithms including naive bayesian tree classifier and sequential minimal optimization was applied on three different data sets of 100 extracted features of the heart sound. The extensive experimental results on the heat sound signals data set demonstrate that the proposed approach outperforms other classifiers and providing the highest classification accuracy with minimized number of features.