Ali Chekima

Comparison of hand segmentation methodologies for Hand gesture recognition

The Hand gesture has provided significant means of communication in human daily interaction and has been widely explored in the Human-Computer Interaction (HCI) studies. This paper presents a comparison of hand segmentation methodologies in the early stage development of an appearance-based hand gesture recognition system for sign language application. The hand segmentation method is based on Jones and Rehg generic skin color model and frame differencing technique that utilized the color and motion cues of image content. Several issues and challenges occurred during the experiments are also discussed and later tackled by the current approach. The present approach applied the idea of integrating both color and motion cues into a single probability map and yield robust features for further tasks. It is merely a first step progress towards an independent environment and signer for the hand gesture recognition system.

Lip detection by the use of neural networks

Lip detection is used in many applications such as face detection and lip reading. In this article, a method for lip detection in color images in a normalized RGB color scheme is presented. In this method, MLP neural networks are used to perform lip detection on segmented skin regions. Several combinations of chrominance components of the normalized RGB color space were used as the input to the neural networks. Two methods were used for obtaining the normalized RGB components from the RGB color scheme. These are called the maximum and intensity normalization methods, respectively. The method was tested on two Asian databases. The number of neurons in the hidden layer was determined by using a modified network-growing algorithm. It was found that the pixel intensity normalization method gave lower lip detection error than the maximum intensity normalization method regardless of the database used, and for most of the combinations of chrominance components. In addition, the combination of the g and r/g chrominance components gave the lowest lip detection error when the pixel intensity normalization method was used for both databases. The effects of the scale and facial expression on lip detection was also studied. It was found that the lip detection error decreased as the scale factor increased. As for facial expression, a laughing facial expression gave the highest lip detection error, followed by smiling and neutral expressions.

Finding the number of hidden neurons for an MLP neural network using coarse to fine search technique

Skin detection is an important preliminary process for subsequent feature extraction in image processing techniques. There are several techniques that are used for skin detection. In this work, the multi-layer perceptron (MLP) neural network is used. One of the important aspects of MLP is how to determine the network topology. The number of neurons in the inputs and output layers are determined by the number of available inputs and required outputs respectively. Thus, the only thing remaining is how to determine the number of neurons in the hidden layer. Therefore, we employed the coarse to fine search method to find the number of neurons. First, the number of hidden neurons is initially set using the binary search mode, HN=1, 2, 4, 8, 16, 32, 64 and 128, where HN indicates the number of hidden neurons. The 30 networks with these HN values are trained and their Mean Squared Error (MSE) is calculated. Then a sequential search, fine search, will be used in the neighbourhood of the HN that gave the lowest MSE. The selected number of neurons in the hidden layer is the lowest HN that gave the lowest MSE. The YCbCr colour space is used in this work due to its capability to separate the luminance and chrominance components explicitly. Several chrominance components are investigated.

Fingerprint identification and recognition using backpropagation neural network

Biometrics is a technology which identifies a person based on his physiology or behavioral characteristics. Fingerprint identification and recognition is a biometrics method that has been widely used in various applications because of its reliability and accuracy in the process of recognizing and verifying a persons identity. The main purpose of this paper is to develop a fingerprint identification and recognition system. The system consists of three main parts, image acquisition, processing and identification and recognition. Fingerprint images are acquired and stored in the database in the image acquisition stage. These images are then enhanced in the image processing stage by performing gray level enhancement, spatial filtering, image sharpening, edge detection, segmentation, and thinning processes. After the image has been processed, it is fed into the backpropagation neural network as input in order to train the network. After training, the neural network is ready to perform the identification and recognition operations (matching process). A neural network has been successfully developed to identify and recognize the core part of fingerprint images.

Palmprint Identification Using Sequential Modified Haar Wavelet Energy

Palmprint identification is the measurement of palmprint features for recognizing the identity of a user. Palmprint is universal, easy to capture and does not change much across time. Palmprint biometric system does not requires specialized acquisition devices. It is user-friendly and more acceptable by the public. Besides that, palmprint contains different types of features, such as geometry features, line features, point features, statistical features and texture features. In this work, peg-less right hand images for 100 different individuals were acquired ten times. No special lighting is used in this setup. The hand image is segmented and its key points are located. The hand image is aligned and cropped according to the key points. The palmprint image is enhanced and resized. Sequential modified Haar transform [1] is applied to the resized palmprint image to obtain modified haar energy (MHE) feature. The sequential modified Haar wavelet can maps the integer-valued signals onto integer-valued signals without abandoning the property of perfect reconstruction. The MHE feature is compared with the feature vectors stored in the database using Euclidean Distance. The accuracy of the MHE feature and Haar energy feature under different decomposition levels and combinations are compared. 94.3678 percent accuracy can be achieved using proposed MHE feature.

Preliminary study of Block Matching Algorithm (BMA) for video coding

The model of advance video coding technique can be divided into two main parts that are spatial model and temporal model. Spatial model exploit the redundancy in its single video frame (I frame) while temporal model exploit the redundancy among frames (P frame). Temporal model deals with motion estimation (ME) and motion compensation (MC) algorithm with the matching technique called “Block Matching Algorithm” (BMA) to produce the next encoded video frame with motion vector. In this paper, seven types of famous BMA technique such as Exhaustive Search, Three Step Search, New Three Step Search, Simple and Efficient Three Step Search, Four Step Search, Diamond Search and Adaptive Rood Pattern Search have been used to analyze the video frames quality with different block size and different sequence of I and P frame.

Data fusion for skin detection

Two methods of data fusion to improve the performance of skin detection were tested. The first method fuses two chrominance components from the same color space, while the second method fuses the outputs of two skin detection methods each based on a different color space. The color spaces used are the normalized red, green, blue (RGB) color space, referred to here as pixel intensity normalization, and a new method of obtaining the R, G, and B components of the normalized RGB color space called maximum intensity normalization. The multilayer perceptron (MLP) neural network and histogram thresholding were used for skin detection. It was found that fusion of two chrominance components gives a lower skin detection error than a single chrominance component regardless of the database or the color space for both skin detection methods. In addition, the fusion of the outputs of two skin detection methods further reduces the skin detection error.

Binary Genetic Algorithm Assisted Multiuser Detector for STBC MC-CDMA

In this paper we investigate the performance of multiuser detector (MUD) for space time block coded (STBC) multi-carrier code-division multiple-access (MC-CDMA) using binary genetic algorithm (GA). Optimum detection for CDMA based systems, including STBC MC-CDMA, using conventional MUD has a high computational complexity. For K number of users for instance, the computational complexity grows exponentially with K. Although suboptimum detection methods with less complexity are available, they are less powerful than MUD. To achieve the optimality of MUD while avoiding the computational complexity, we propose the use of binary genetic algorithm in the detector. For the evaluation, we simulated STBC MC-CDMA system over a frequency selective rayleigh fading channel. The receiver employs the binary genetic algorithm assisted MUD. For the purpose of comparison, STBC MC-CDMA receiver using equal gain combining (EGC) was simulated as well. The simulation result shows that the system under investigation performed better than STBC MC-CDMA using EGC.

Probabilistic multi SVM weed species classification for weed scouting and selective spot weeding

In this paper, a probabilistic output multi SVMs were used to classify the weed seedlings into groups for spot spraying and weed scouting application. Weeds in the samples are collected at approximely 1-4 weeks after post emergence. The weed seedlings are classified using Support Vector machines while feature selection and fine tuning of classifier parameters were fine tuned using genetic algorithm. The features which included regional shapes parameters, fractal dimensions and elliptical Fourier coefficients, skeleton statistics, boundary to centroid and colour statistics were extracted from individual leaves and the overall binarized shape of the weed seedlings. The resulting SVM ensemble classifier is able to classify the various weed seedlings into various classes at a reasonable rate which can be further improved by enlarging training sets and improving individual SVMs.

Hybrid Component-Based Face Recognition System

Face recognition system is a fast growing research field because of its potential as an eminent tool for security surveillance, human-computer interaction, identification declaration and other applications. Face recognition techniques can be categorized into 3 categories namely holistic approach, feature-based approach, and hybrid approach. In this paper, a hybrid component-based system is proposed. Linear discriminant analysis (LDA) is used to extract the feature from each component. The outputs from the individual components are then combined to give the final recognition output. Two methods are used to obtain the components, namely the facial landmarks and the sub-images. It was found out that the fusion of the components does improve the recognition rate compared to individual results of each component. From the sub-image method, it can be seen that as the size of the components get smaller, the recognition rate tends increase but not always.