Amine Nait-Ali

An input-delay neural-network-based approach for piecewise ECG signal compression

We propose an input delay neural network (IDNN) based time series prediction algorithm for compressing electrocardiogram (ECG) signals. Our algorithm has been tested and successfully compared vis-a/spl grave/-vis other popular techniques for its compression efficiency and reconstruction capability.

An adaptive calibration of an infrared light device used for gaze tracking

In this study, the measurement of gaze position is used to indicate the areas that attract the subjects attention in an image. An infrared light device is used to provide the horizontal and vertical eye movement. The calibration procedure is studied to perform the mapping between equipment coordinates and the image coordinates. A polynomial transformation of higher order is used to model this mapping by using a mean square error criterion. The method developed here allows an adaptive correction of the measured gaze positon according to users characteristics.

A Postural Information-Based Biometric Authentication System Employing S-Transform, Radial Basis Function Network, and Extended Kalman Filtering

This paper proposes a new system for biometry-based human authentication, where postural signal information is utilized to identify a person. The system employs a novel approach where four types of temporal postural signals are acquired for each person to develop an authentication database, and for each posture, both signals in the - and -directions are utilized for the purpose of authentication. The proposed system utilizes S-transform, which is a joint time-frequency representation tool, to determine the characteristic features for each human posture. Based on these characteristic features, a radial basis function network (RBFN) system is developed for the purpose of specific authentication. The RBFN authentication system is developed by training it to employ extended Kalman filtering (EKF). The EKF-trained RBFN authentication system could produce overall authentication accuracy on the order of 94%-95% and could outperform similar authentication systems developed, which employ two very popular variants of backpropagation neural networks (BPNNs) and a variant of radial basis neural network (RBNN).

ECG compression method using Lorentzian functions model

An ECG compression algorithm using a combination of Lorentzian functions model is proposed in this paper. In order to estimate the parameters of the Lorentzian functions, the discrete Fourier transform (DFT) is first applied to a mean removed ECG signal from which only the most significant DFT coefficients are retained. The obtained coefficients are, then modeled as the sum of a given number of superimposed exponentially damped sinusoids (EDS), commonly identified by their amplitudes, real damping factors, frequencies and initial phases. Finally, these EDS parameters are estimated, using SVD method, then coded. The algorithm has been tested for its coding efficiency and reconstruction capability by applying it to several popular, benchmark ECG signals. Encouraging results have been obtained.

Multivariate statistical modeling of images in sparse multiscale transforms domain

In this paper, we propose a multivariate statistical model to characterize the inter- and intra-scale dependencies between image coefficients in the oriented and non-oriented sparse multiscale transforms domain. Our proposed model, namely the Multivariate Bessel K Form, is based on multivariate extension of Bessel K Form distribution. To establish this model in practice, we propose an analytical form of PDF and then estimate its hyperparameters. Also, we compared it to the other models proposed in literature such as the Anisotropic Multivariate Generalized Gaussian and the Jeffrey models, in order to demonstrate its capabilities to capture the inter- and intra-scale dependencies between image detail coefficients.

Fast communication: Generalized multi-directional discrete Radon transform

This paper presents a discrete generalized multi-directional Radon transform (GMDRT) and its exact inversion algorithm. GMDRT is an extension of the classical Radon transform. It aims to project parameterized curves and geometric objects following several directions. For this purpose, we propose an algebraic formalism of the Radon Transform presenting the forward transform as a matrix-vector_ multiplication. We show in this paper that the exact inversion of the GMDRT exists. This property allows useful applications, in the field of digital image processing.

Hidden biometrics: Towards using biosignals and biomedical images for security applications

When dealing with biometrics, we generally refer to security biometrics which is a set of techniques used to identify an individual using his biological or behavioral features. But sometimes, biometrics, in particular medical biometrics, refers to some specific methods that are used to quantify or to measure some parameters extracted from medical data. In this paper, we bridge the gap between the security biometrics and the medical biometrics and we try to discuss and highlight the idea which consists in using medical data, such as biosignals, MRI images and X-Ray images for the purpose of individual identification or verification. This is what we call the “Hidden biometrics” or “Intrinsic biometrics”. As we will see, some of the techniques using biosignals are suited for applications requiring frequent up-dates and other approaches which use medical images are particularly robust regarding any potential forgery.

A novel approach based brain biometrics: Some preliminary results for individual identification

Numerous anatomical studies of the human brain have shown a significant inter-individual variability of brain characteristics. Specifically, the extracted characteristics are used in our application as a biometric tool to identify individuals. For this purpose, Magnetic Resonance Imaging (MRI) images are considered. We show that using a single slice from an MRI volumetric image, acquired at a given level, one can extract significant brain codes that can be used for the purpose to identify individuals. Explicitly, the proposed biometric approach uses some coding techniques that are commonly employed for iris identification. Specifically, 1D log Gabor Wavelet has been considered for feature extraction. Finally, the proposed algorithm is evaluated on the Open Access Series of Imaging Studies (OASIS) database containing brain MRI Images. Results using 210 classes show that high accuracy of 98.25% to identify individuals are obtained.

ECG Compression Using an Ensemble Polynomial Modeling: Comparison with the DCT Based Technique

The proposed ECG compression technique combined two approaches, ECG beat alignment and the polynomial modelling. QRS complexes are first detected then aligned in order to reduce high frequency changes from beat to beat. These changes are modelled by means of a polynomial projection. ECGs from MIT-BIH database are used to evaluate the performance of the proposed technique. A comparison with the DCT approach is performed by means CR/PRD curves.

Application of simulated annealing for estimating BAEPs in endocochlear pathologies

In this paper, we propose a new approach aimed at handling the temporal Brainstem Auditory Evoked Potentials (BAEPs) non-stationarity. It is pointed out that for some endocochlear pathologies, BAEPs could be randomly delayed from one response to another. This non-stationarity leads to smoothed BAEPs when applying ensemble averaging or any other technique based on BAEPs stationarity. In that case, waves identification is very difficult, sometimes impossible. The problem consists in estimating time delays. Knowing the distribution of delays allows subsequent study of the dynamic of the cochlea and, perhaps, identification of the nature of its pathology. The approach suggested in this paper is based on Simulated Annealing, used to minimize a non-linear criterion involving delays. This technique is advantageously compared to the non-corrected ensemble averaging method, using a set of simulated data based on a realistic model. As an illustration, results based on real signals recorded from two patients are presented and discussed at the end of the paper.