Alaa Sagheer

Face recognition across illumination

Illumination variation on images of faces is one of the most difficult problems in face recognition systems. The performance of a self-organizing map-based face recognition system is highly degraded when the illumination in test images differs from that of the training images. Illumination normalization is a way to solve this problem. Both global and local image enhancement methods are studied in this article. A local histogram equalization method strongly improves the recognition accuracy of the CMU-PIE face database.

Visual speech features representation for automatic lip-reading

A fundamental task in the pattern recognition field is to find a suitable representation for a feature. We present a new visual speech feature representation approach that combines hypercolumn model (HCM) with HMM to perform a complete lip-reading system. We use HCM to extract visual speech features from the input image. The extracted features are modeled by Gaussian distributions using HMM. The proposed lip-reading system can work under varying lip positions and sizes. All images were captured in a natural environment without using special lighting or lip markers. Experimental results are shown to compare favourably with the results of two reported systems, SOM and DCT based systems. HCM provides better performance than both of these systems.

An Effective Face Detection Algorithm Based on Skin Color Information

Face detection approach is presented in this paper combines skin color detection and neural network. The first motivation for our paper is to decide which color space is the best in order to build efficient skin color detector can be embedded in the overall face detection system. The proposed skin detection approach uses a chrominance distribution model of skin-color information in the input image in order to detect skin pixels over the entire image. Next, morphological operations are used in order to smooth the detected skin region and generate, finally, face candidates for face-base applications. Finally, neural network is used in order to verify these face candidates. Many experiments using color images gathered from the Internet and from our own database are conducted and give encouraging results. It is expected to combine the proposed face detector with face recognition approach to be embedded later in human computer interaction applications.

Hyper column model vs. fast DCT for feature extraction in visual Arabic speech recognition

Recently, the multimedia signal processing community has shown increasing interest for research development on visual speech recognition domain. In this paper we present a novel visual speech recognition approach based on our model hyper column model (HCM). HCM is used for feature extraction task. The extracted features are modeled by Gaussian distributions through using hidden Markov model (HMM). The proposed system, HCM and HMM, can be used for any visual recognition task. We use it here to comprise a complete lip-reading system and evaluate its performance using Arabic database set. According to our knowledge, this is the first time that visual speech recognition is applied for Arabic language. Toward fair evaluation we compare our accuracy results with those using fast discrete cosine transform (FDCT) approach, in a separate experiment and using same data set and conditions of HCM experiment. Comparison turns out that HCM shows higher recognition accuracy than FDCT for Arabic sentences and words. HCM does not provide higher accuracy only but also it capable to achieve shift invariant recognition whereas FDCT can not

Self-Organizing Feature Maps for HMM Based Lip-Reading

Audio-visual dialogue is an appealing tool for natural interface with computers. Lip-reading is one of important part for audio-visual dialogue. In this paper, it is proposed to use a self-organizing feature map (SOM) and a hierarchical SOM: Hypercolumn model (HCM), as a module of phoneme feature space construction for HMM base lip-reading system. Those SOMs allow alleviating many difficulties associated with feature space construction. It is, however, required for on-line systems to reduce the feature extraction time to the range of normal video camera rates. To achieve this, a randomization technique is introduced. The experimental results show performances of the SOMs for Japanese lip-reading.

Quantum coding in non-inertial frames

The quantum and classical correlations are quantified by means of the coherent and mutual information, respectively, where we use the single-mode approximation. It is shown that the users can communicate in an optimal way for small values of accelerations. The capacity of accelerated channel is investigated for different classes of initial states. It is shown that the capacities of the traveling channels depend on the frame in which the accelerated channels are observed in and the initial shared state between the partners. In some frames, the capacities decay as the accelerations of both qubit increase. The decay rate is larger if the partners initially share a maximum entangled state. The possibility of using the accelerated quantum channels to perform quantum coding protocol is discussed. The amount of decoded information is quantified for different cases, where it decays as the partner’s accelerations increase to reach its minimum bound. This minimum bound depends on the initial shared states, and it is large for maximum entangled state.

Piecewise one dimensional Self Organizing Map for fast feature extraction

It is well known that the problem arising from high dimensionality of data should be considered in pattern recognition field. Face recognition databases are usually high dimensionality, especially when limited training samples are available for each subject. Traditional techniques perform dimensionality reduction are unable to solve this problem smoothly, which makes feature extraction task much difficult. As such, a novel method performs feature extraction and dimensionality reduction for high-dimensional data is needed. In this paper, a new algorithm for traditional Self Organizing Map (SOM) is presented to cope with this problem with low computation cost. It is shown here that the computation cost of the proposed approach, comparing to traditional SOM is reduced into O(d<inf>1</inf>+ d<inf>2</inf> +…+ d<inf>N</inf>) instead of O(d<inf>1</inf> × d<inf>2</inf> ×… × d<inf>N</inf>), where d<inf>j</inf> is the number of neurons through a dimension d<inf>j</inf> of the feature map. Experiments are carried out using benchmark database show that the proposed algorithm is a good alternate to traditional SOM, especially, when high-dimensional feature space is desired.

Communication via quantum neural networks

In this paper, a quantum teleportation protocol based on quantum neural networks is presented. We studied the relation between the networks weight and the fidelity of transferring information among neurons. It has been found that as the networks weight increases the accuracy of the transformed information increases. Preliminary results of a practical example are given in this paper.

Improved SOM search algorithm for high-dimensional data with application to face recognition across pose and illumination

In this paper we focus on dealing with large size databases. Such databases require the construction of suitable feature spaces to accommodate data. The paper presents a new search algorithm based on the self organizing map (SOM) avoids the high-cost of computation in such cases. The proposed SOM algorithm is combined with support vector machine (SVM) to form a new appearance based approach. The proposed approach is evaluated in face recognition experiments across variations in pose and illumination. A huge-size database is used to judge effectively the proposed approach. The results have compared with another reported approach based on light field theory using same huge database.

An autonomous competitive learning algorithm using quantum hamming neural networks

Quantum neural network is a fledging research domain based on the merge of classical neural network and quantum computing. In this paper, we add the quantum effect to the classical hamming neural network algorithm in order to employ the advantages of quantum information to yield, finally, a novel quantum competitive learning algorithm. The proposed algorithm, called quantum hamming neural network (QHNN), is capable to recognize incomplete patterns, as well as, increase the probability of recognizing patterns on the account of undesired patterns. Moreover, these undesired patterns could be used as new patterns for training the algorithm in subsequent steps. The proposed algorithm is testified via a case study and a classification experiment, where promising results, reaches to 100%, are given and compared favorably with other reported quantum competitive algorithms.