Mohamed I. Alkanhal

Automatic Stochastic Arabic Spelling Correction With Emphasis on Space Insertions and Deletions

This paper presents a stochastic-based approach for misspelling correction of Arabic text. In this approach, a context-based two-layer system is utilized to automatically correct misspelled words in large datasets. The first layer produces a list in which possible alternatives for each misspelled word are ranked using the Damerau-Levenshtein edit distance. The same layer also considers merged and split words resulting from deletion and insertion of space character. The right alternative for each misspelled word is stochastically selected based on the maximum marginal probability via A* lattice search and m-gram probability estimation. A large dataset was utilized to build and test the system. The testing results show that as we increase the size of the training set, the performance improves reaching 97.9% of F1 score for detection and 92.3% of F1 score for correction.

A Manual System to Segment and Transcribe Arabic Speech

In this paper, we present our first work in the ¿computerized teaching of the Holly Quran¿ project, which aims to assist the memorization process of the Noble Quran based-on the speech recognition techniques. In order to build a high performance speech recognition system for this purpose, accurate acoustic models are essentials. Since annotated speech corpus of the Quranic sounds was not available yet, we tried to collect speech data from reciters memorizing the Quran and then focusing on their labeling and segmentation. It was necessarily, to propose a new labeling scheme which is able to cover all the Quranic Sounds and its phonological variations. In this paper, we present a set of labels that cover all the Arabic phonemes and their allophones and then show how it can be efficiently used to segment our Quranic corpus.

Speaker verification based on Saudi accented Arabic database

Speaker verification is concerned with verifying the speakerpsilas claimed identity. This paper reports on recent experiments we carried out for speaker verification using a Saudi accented Arabic telephone speech database with 1033 speakers. Gaussian Mixture Model was employed in these experiments. In speaker verification, users might produce two or more utterances. We show that we can reduce error rates by combining scores of these utterances.

Multispectral Periocular Classification With Multimodal Compact Multi-Linear Pooling

Feature-level fusion approaches for multispectral biometrics are mainly grouped into two categories: 1) concatenation and 2) elementwise multiplication. While concatenation of feature vectors has benefits in allowing all elements to interact, it is difficult to learn output classification. Differently, elementwise multiplication has the benefits in enabling multiplicative interaction, but it is difficult to learn input embedding. In this paper, we propose a novel approach to combine the benefits of both categories based on a compact representation of two feature vectors’ outer product, which is called the multimodal compact multi-linear pooling technique. We first propose to expand the bilinear pooling technique for two inputs to a multi-linear technique to accommodate for multiple inputs (multiple inputs from multiple spectra are frequent in the multispectral biometric context). This fusion approach not only allows all elements to interact and enables multiplicative interaction, but also uses a small number of parameters and low computation complexity. Based on this fusion proposal, we subsequently propose a complete multispectral periocular recognition system. Employing higher order spectra features with an elliptical sampling approach proposed by Algashaam et al., our proposed system achieves the state-of-the-art performance in both our own and the IIIT multispectral periocular data sets. The proposed approach can also be extended to other biometric modalities.

A robust recognition system for partially occluded faces

Correlation filters have shown good performance results for distortion tolerant applications especially in target and face recognition problems. In this paper, we investigate the performance of these filters when applied to partially occluded human faces. We present a system for eye region recognition based on a special class of unconstrained correlation filters called optimal trade off Maximum Average Correlation Height (OT-MACH) filter. This system is useful for people who cover their faces, due to, for example, diseases or cultural reasons. The performance of this system is evaluated using the extended Yale B dataset. Our experimental results show that this system is robust to occlusion compared to the principal component analysis (PCA) and the local binary pattern (LBP). The OT-MACH filter shows error rates of 0.31% and 10.31% for non-occluded and occluded face recognition systems, respectively.

A hybrid automatic scoring system for Arabic essays

Essay writing is widely used for student performance assessment. This paper presents a hybrid automatic essay scoring system AES for Arabic essays. The system attempts at saving the time teachers spend on reading and scoring Arabic essays. It utilizes latent semantic analysis LSA and three linguistic features i.e., word stemming, number of words and number of spelling mistakes. This paper also describes an algorithm to determine the optimal reduced dimensionality used in LSA. To evaluate the performance of this system, an Arabic dataset was developed based on essays collected from college students. The experimental results show the effectiveness of using LSA for scoring Arabic essays, especially when combined with other linguistic features. The system shows that 96.72% of the test data are correctly scored and the correlation between automatic and manual scores is 0.78, which is close to the interhuman correlation of 0.7.

Mubser: a bilingual Braille to text translation with an Arabic interface

Purpose – The purpose of this paper is to describe the development of a system called Mubser to translate Arabic and English Braille into normal text. The system can automatically detect the source language and the Braille grade.Design/methodology/approach – Mubser system was designed under the MS‐Windows environment and implemented using Visual C# 2.0 with an Arabic interface. The system uses the concept of rule file to translate supported languages from Braille to text. The rule file is based on XML format. The identification of the source language and grade is based on a statistical approach.Findings – From the literature review, the authors found that most researches and products do not support bilingual translation from Braille to text in either contracted or un‐contracted Braille. Mubser system is a robust system that fills that gap. It helps both visually impaired and sighted people, especially Arabic native speakers, to translate from Braille to text.Research limitations/implications – Mubser is b...

Masked correlation filters for partially occluded face recognition

Face recognition is widely used for a variety of applications, such as identifying people for security purposes, as well as photo album organization. A challenge is to perform accurate face recognition when there exist partial occlusions of the face such as scarves or sunglasses. Correlation Filters (CFs) are an occlusion-tolerant object recognition method, potentially suited to deal with partial occlusions. In this paper, we introduce a new class of correlation filters called Masked Correlation Filters (MCFs), that are designed specifically to handle partial occlusions in face images. The benefits of using MCFs are illustrated using well-known face image data sets.

Polynomial Correlation Filters for Human Face Recognition

This paper describes a nonlinear face recognition method based on polynomial spatial frequency image processing. This nonlinear method is known as the polynomial distance classifier correlation filter (PDCCF). PDCCF is a member of a well-known family of filters called correlation filters. Correlation filters are attractive because of their shift invariance and potential for distortion tolerant pattern recognition. PDCCF addresses more than one filter in the system, each one with a different form of non-linearity. Our experimental results on the Olivetti Research Laboratory (ORL) and Extended Yale B (EYB) face datasets show that PDCCF outperforms the principal component analysis (PCA), and the local binary pattern (LBP).

Speech Rating System through Space Mapping

Predicting human behavior has been the subject of many research areas especially in machine learning. Due to its potential benefits, financially or otherwise, researchers have focused on modeling human behavior from recommending items in an online store to predicting the behavior of an entire ecosystem. In this paper, we make an attempt to predict human preference towards natural speech. The proposed approach makes use of extracted user features from the dataset using Singular Value Decomposition (SVD), features extracted from the wave signal using Mel-frequency cepstral coefficients (MFCC) and Radial Basis Function to map the two feature-spaces. The proposed approach was able to reach a Pearson Correlation Coefficient of 0.92 and a 0.258 MAE when compared to the original average scores. The main contribution of the presented work is the fact that mapping the signal-features (MFCC) into an intermediate feature space (SVD) is far more effective than mapping the signal features directly into the desired output. The proposed algorithm outperformed Support Vector Machines (SVM) in all measures, precisely by 88.14% in terms of correlation and by 48.62% in terms of error.