Prof. Dr. Amine NAIT-ALI

E. Farazdaghi, Thesis in progress

Facial ageing modelling consists in  two parts: Make someone older and make an adult younger. Making someone younger is more difficult and challenging part of the face modelling because it needs to change both craniofacial morphology and skin texture. Now we are done with this part. We propose a component based method for introducing a geometrical model to show an adult face in the childhood. In this recursive model, the face contour and components are changed non-linearly.

We are using an adult facial image as an input, facial components and contour and are changed according to extracted measures belong to the target cluster. Final results come from merging the geometrical result and face mean in the same cluster. Using this approach, we can make the individual younger, up to 3 years old.

Supervisor: Prof. A. Nait-ali.

D. Kerbaj, Thesis in progress

The main work was focused on a defined protocol with the objective to determine authentication methods using brain waves. For this protocol, I did not have any participating subjects since the protocol was applied to me. EEG brain signals were recorded (from my scalp) while forcing thoughts, and/or imaginary movements in order to compare multiple acquisition recording leading to identify a repetitive pattern to be able to authenticate myself later on from simple analysis of my brain waves. Results will be published in an article.

Supervisors: Prof. A. Nait-ali and Prof. W. Hassan.

I, El Ouadi, Thesis in progress

In this theoretical study, Discrete Generalized Multi Directions Radon Transform (GMDRT) and its exact inversion algorithm is considered as a pattern recognition tool. 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 have proposed an algebraic formalism of Radon Transform presenting the forward transform as a matrix-vector multiplication. We showed that the exact inversion of the GMDRT exists. This property allows useful applications within digital image processing field.

Supervisors: Prof. A. Nait-ali, Co-supervisors: Dr. R. Fournier and Prof. K. Bsais/Dr. A. Hamouda.

Y. Kabbara, 2015 (Thesis in french)

Palm-print based biometric systems are known to be sensitive to spoofing. The idea behind this thesis is to propose some new approaches to identify persons using hand X-ray images. Even if such systems cannot be used in daily routine, this technique can be employed when high-level security is required. In terms of image processing, a segmentation technique is used to extract regions of interest; therefore each region is modeled using a certain number of parameters.These parameters are considered as a signature for each individual.

Supervisors: Prof. A. Nait-ali. Co-supervisors: Dr. A. Shahin and Prof. M. Khalil.

D. Maatar, 2013 (Thesis in french)

The analyzed approach consists in using the stabilogram signal as a soft biometric tool in order to extract some relevant features from a standing position of persons. Using the PCA (Principal Analysis Component) decomposition method, center of pressure time series are analyzed. More specifically, the stabilogram is decomposed into three components, namely: trend, rambling and trembling. Studying the trace of analytic trembling (respectively of rambling) in the complex plan highlights a unique rotation center. This specification allows the definition of the phase and allows the extraction of phase fluctuation. Adapting the stabilogram diffusion analysis method (SDA), Hurst exponents (H1 and H2) are extracted from the diffusion of phase’s fluctuations related either to trembling and rambling. These parameters represent efficient indicators related to postural equilibrium status of healthy subjects.

Supervisors: Prof. A. Nait-ali, Co-supervisors: Dr. R. Fournier and Prof. Z. Lachiri.

K. Aloui, 2012 (Thesis in french)

The purpose of this work is to study an approach which consists in using MRI images for the identification or verification of persons. In other words, we ask the following question: can one identify individuals using their brain geometry characteristics? Our aim is to validate the feasibility of this new biometric tool based on human brain characterization. The proposed approach differs from existent biometrics modalities (e.g. fingerprint, hand, etc) in the sense that brain features cannot be spoofed as it is the case when dealing with fake fingerprints, or fake hands. In this work, we consider volumetric Magnetic Resonance Images (MRI) from which brain shapes are extracted using a 3D level-sets segmentation approach. Afterwards, geometrical descriptors are extracted from the 3D brain volume and from a projected version which provide specific features such as the isoperimetric ratio, the cortical surface curvature and the Gyrification index. Brain characteristics can also be extracted from 2D slices In such a scenario, we have proposed an approach similar to the one used for Iris recognition (Daugman algorithm).

Supervisors: Prof. A. Nait-ali and Prof. S. Naceur.

A. Bouchemha, 2012 (Thesis in french)

The proposed technique extracts palmprint features using jointly, Radon transform and a geometric Delaunay triangulation. In such a process, Radon transform allows to extract directional characteristics from the palm of the hand. Afterwards, the most significant information is structured by achieving Delaunay triangulation, providing hence, a specific signature of the palmprint. In order to compare the uniqueness as well as the stability of the palmprint signature, Hausdorff distance has been used as a criterion of similarity.

Supervisors: Prof. A. Nait-ali and Prof. N. Doghmane.

Thesis 1: S. Chantaf, 2011 (Thesis in french), Thesis 2: N. Belgacem, 2014 (Thesis in french)

We have been interested in our research group to human identification using normal ECG signals (non-pathological cases are considered). Only the most significant ECG parameters extracted from recorded signals are considered. Specifically, these parameters are extracted by modelling the ECG using wavelet networks. The radial basis neural network method is then used to classify these parameters. Consequently, a useful analysis is performed to evaluate the robustness of the identification in some specific situations such as: noise presence, heart variability and electrode displacements. This work required two realted thesis. 

Thesis 1: Supervisors: Prof. A. Nait-ali and Prof. M. Khalil.

Thesis 2: Supervisors: Prof. A. Nait-ali and Prof. Breksi.

Thesis 1: E. Zeybek, 2011 (Thesis in french), Thesis 2: T. Brahimi (Thesis in French)

Multimodal Compression (MC) is a new approach that we have developed in order to encode jointly data, acquired from numerous modalities. For example, one can encode jointly and image and a set of signals. Generally speaking, signal samples are merged in the spatial domain of the image, by avoiding the Region Of Interest (ROI) ,using a specific mixing function (not to be confused with watermarking). Afterwards, the whole mixture is compressed using any standard such as JPEG 2000. The spatial mixing function inserts samples in a selected region after a down-sampling process. The decoding is achieved by inverting the process using a separation function. Results show that this technique allows better performances in terms of Compression Ratio (CR) compared to approaches which encode separately modalities. This apporach has been evaluated for medical data and extended to other types of image and signals. This topic was studies through two thesis. 

Thesis 1: Supervisors: Prof. Amine Nait-ali, co-supervisor: Dr. R. Fournier.

Thesis 2: Supervisors: Prof. Amine Nait-ali, co-supervisor: Dr. R. Fournier.