Abstract In this work we propose a hybrid NN/HMM model for online Arabic handwriting recognition. The proposed system is based on Hidden Markov Models (HMMs) and Multi Layer Perceptron Neural Networks (MLPNNs). The input signal is segmented to continuous strokes called segments based on the Beta-Elliptical strategy by inspecting the extremum points of the curvilinear velocity profile. A neural network trained... In this work we propose a hybrid NN/HMM model for online Arabic handwriting recognition. The proposed system is based on Hidden Markov Models (HMMs) and Multi Layer Perceptron Neural Networks (MLPNNs). The input signal is segmented to continuous strokes called segments based on the Beta-Elliptical strategy by inspecting the extremum points of the curvilinear velocity profile. A neural network trained with segment level contextual information is used to extract class character probabilities. The output of this network is decoded by HMMs to provide character level recognition. In evaluations on the ADAB database, we achieved 96.4% character recognition accuracy that is statistically significantly important in comparison with character recognition accuracies obtained from state-of-the-art online Arabic systems. Read More Read Less

Abstract Vector quantization based on a codebook is a basic process to recognition the speech signal by discrete hidden markov model. This article identifies the fundamental importance of vector quantization codebooks in the performance of the system. For comparison, two different algorithms k-means and Gaussian mixture models (GMM) have been used to obtain two sets of speech feature codebook. We used... Vector quantization based on a codebook is a basic process to recognition the speech signal by discrete hidden markov model. This article identifies the fundamental importance of vector quantization codebooks in the performance of the system. For comparison, two different algorithms k-means and Gaussian mixture models (GMM) have been used to obtain two sets of speech feature codebook. We used in analysis phase Mel Frequency Cepstral Coefficients (MFCC) supplement by dynamic features to increase system performance, although experiments are carried out for the choice of the optimal parameters of the system. Good results are obtained using a GMM method. Read More Read Less

Abstract All natural languages exhibit a great deal of internal variations in terms of a specific set of linguistic items or human speech patterns such as sounds, words or grammatical features which can uniquely associate with some external sociolinguistic factors such as geographical area or social group. The present study investigated the probabilities of occurrence of seventeen function words, thirteen content... All natural languages exhibit a great deal of internal variations in terms of a specific set of linguistic items or human speech patterns such as sounds, words or grammatical features which can uniquely associate with some external sociolinguistic factors such as geographical area or social group. The present study investigated the probabilities of occurrence of seventeen function words, thirteen content words and three idiosyncratic features in the electronic texts worded in the English language variants from Cameroon and four West African countries, namely Ghana, Liberia, Nigeria and Sierra Leone. The findings show that there is no significant difference in the distribution of the word ‘the’ among all corpora. Multiple comparison tests revealed different levels of relationships among the countries’ corpora. The study reveals linguistic similarities between the Sierra-Leonean and Liberian texts and between the Ghanaian and Nigerian texts, while Cameroonian corpus was found to be the most significantly different from that of Nigeria. Read More Read Less

Abstract Vector quantization based on a codebook is a basic process to recognition the speech signal by discrete hidden markov model. This article identifies the fundamental importance of vector quantization codebooks in the performance of the system. For comparison, two different algorithms k-means and Gaussian mixture models (GMM) have been used to obtain two sets of speech feature codebook. We used... Vector quantization based on a codebook is a basic process to recognition the speech signal by discrete hidden markov model. This article identifies the fundamental importance of vector quantization codebooks in the performance of the system. For comparison, two different algorithms k-means and Gaussian mixture models (GMM) have been used to obtain two sets of speech feature codebook. We used in analysis phase Mel Frequency Cepstral Coefficients (MFCC) supplement by dynamic features to increase system performance, although experiments are carried out for the choice of the optimal parameters of the system. Good results are obtained using a GMM method. Read More Read Less

Abstract All natural languages exhibit a great deal of internal variations in terms of a specific set of linguistic items or human speech patterns such as sounds, words or grammatical features which can uniquely associate with some external sociolinguistic factors such as geographical area or social group. The present study investigated the probabilities of occurrence of seventeen function words, thirteen content... All natural languages exhibit a great deal of internal variations in terms of a specific set of linguistic items or human speech patterns such as sounds, words or grammatical features which can uniquely associate with some external sociolinguistic factors such as geographical area or social group. The present study investigated the probabilities of occurrence of seventeen function words, thirteen content words and three idiosyncratic features in the electronic texts worded in the English language variants from Cameroon and four West African countries, namely Ghana, Liberia, Nigeria and Sierra Leone. The findings show that there is no significant difference in the distribution of the word ‘the’ among all corpora. Multiple comparison tests revealed different levels of relationships among the countries’ corpora. The study reveals linguistic similarities between the Sierra-Leonean and Liberian texts and between the Ghanaian and Nigerian texts, while Cameroonian corpus was found to be the most significantly different from that of Nigeria. Read More Read Less