Tarek Frikha

Adaptive architecture for medical application case study: Evoked Potential detection using matching poursuit consensus

The emergency of embedded systems puts new challenges for the design of different system in many fields. One of the embedded applications fields is the medical one. The major difficulty is the embedded systems reduced energy and computational resources that must be carefully used to execute complex application often in unpredictable environments. In this paper, the used application is the detection of evoked potential with variable latency and multiple trials using consensus matching pursuit. Fitting to the noisy Evoked Potential (EP) signal persistent in all response, we use the Consensus version of the matching pursuit algorithm (CMP). EP is a resulted wave from a stimulus. The EP can be explained with a good quality of energy ratio factor (QR). If we use a noisy EP, we cannot reconstruct the original data because of the random atoms of CMP dictionary. We select the significant atoms to rebuild and EP signals. This application is embedded on a Xilinx ML 507. We used an adaptive architecture based on dynamically partial reconfiguration.

Integration of stationary wavelet transform on a dynamic partial reconfiguration for recognition of pre-ictal gamma oscillations

To define the neural networks responsible of an epileptic seizure, it is useful to perform advanced signal processing techniques. In this context, electrophysiological signals present three types of waves: oscillations, spikes, and a mixture of both. Recent studies show that spikes and oscillations should be separated properly in order to define the accurate neural connectivity during the pre-ictal, seizure and inter-ictal states. Retrieving oscillatory activity is a sensitive task due to the frequency overlap between oscillations and transient activities. Advanced filtering techniques have been proposed to ensure a good separation between oscillations and spikes. It would be interesting to apply them in real time for instantaneous monitoring, seizure warning or neurofeedback systems. This requires improving execution time. This constraint can be overcome using embedded systems that combine hardware and software in an optimized architecture. We propose here to implement a stationary wavelet transform (SWT) as an adaptive filtering technique retaining only pre-ictal gamma oscillations, as validated in previous work, on a partial dynamic configuration. Then, the same architecture is used with further modifications to integrate spatio temporal mapping for an early recognition of seizure build-up. Data that contains transient, pre-ictal gamma oscillations and a seizure was simulated. the method on real intracerebral signals was also tested. The SWT was integrated on an embedded architecture. This architecture permits a spatio temporal mapping to detect the accurate time and localization of seizure build-up, while reducing computation time by a factor of around 40. Embedded systems are a promising venue for real-time applications in clinical systems for epilepsy.

Embedded approach for edge recognition: Case study: Vehicle registration plate recognition

The development of multimedia embedded applications continues to increase. As an example, image recognition applications are not only developed for the PCs but also for the embedded systems such as the face recognition, the texture detection and the edge detection. In this paper, we will present a hardware architecture based on an adaptation approach. To validate this architecture, an edge detection approach is chosen. To illustrate the proposed approach, vehicle registration plate recognition is proposed. The latter application was tested in a first time by a software implementation. n a second time, we use the co design technique based on a mixed Hardware Software architecture. The used platform for the implementation is the Xilinx ML 507 FPGA platform.

Use of ridgelets, curvelets application for face recognition: Case study: Smart identity card

This paper presents a brief description of ridglet transform, curvelet transform: the first and second generation, we detail the various areas where these transforms have proven their qualities and their contributions compared to the previous transformed. As case study of these transforms, smart identity cart application using smart watermarking techniques is proposed.

Self adaptive augmented reality systems on FPGA

Augmented reality (AR) systems emerged and become a very gifted 3D embedded multimedia application. AR consists on adding specific 3Ds animations on a video flow. The design and the implementation of such systems on FPGA are complex and difficult. To reduce computational resources that must be carefully used to execute complex application. This execution can be often done in unpredictable environments: it is the major problem to solve. The system architecture must be efficient and flexible enough to adapt system resources to the application requirements and the environment architectures and mobiles constraints. In this paper, we describe our concept of flexible architecture: we have developed IPs to obtain self-adaptive augmented reality systems to implement on FPGA.

Design of an Adaptive 3D Graphics Embedded System

The big expansion of complex multimedia applications particularly in nomad and battery operated systems puts new challenges for system designers. Such systems have small energy and computational resources. They often work in unpredictable environments and run complex data-dependant applications with variable requirements and computational complexity. The system architecture must be energy efficient and powerful to deal with the complexity of supported applications and the system autonomy requirements. On the other hand, the system architecture must be flexible to allow efficient self adaptation. In this paper, we present an adaptation technique for the respect of the system battery lifetime using a best effort strategy for the QdS application constraint and timing constraints. A high level simulator to ease the choice of the proposed adaptation technique is also presented.

A Comparison Between Modeling a Normal and an Epileptic State Using the FHN and the Epileptor Model

In spite of important technological developments in the medical field and particularly in neuroscience one, epilepsy remained a serious pathology that could affect the human brain. In this work, we modeled a healthy and an epileptic cerebral activity in rest state. We used, the virtual brain TVB toolbox to simulate the two states based on FHN and epileptor model. We compared phase plane spaces, electrophysiological time series (electroencephalogram EEG, magnetoencephalogram MEG and intracerabral EEG), specter of eigenvalues transition matrix and topographic maps for healthy and epileptic rest state. There is a unique metastable state for healthy cerebral dynamics convergence which disappears in epileptic cerebral dynamics. Epileptic rest state time series depicts several transitory activities that vanish in the normal state. Normal rest state topographic maps illustrate a limited dipolar activity; which is more extended in epileptic model. These prominent differences would have an important impact on real cerebral activities analysis.

Design of a dynamically reconfigurable architecture for the 3D image synthesis

In this paper, we present a dynamically reconfigurable (DR) architecture for a 3D image synthesis application. We address different issues not covered in similar works especially the use of low complex and cost FPGA and the simultaneous support of different constraints like the energy consumption and the real time constraint. The proposed system uses an adaptation module that monitors the internal architecture modifications using FPGA dynamic reconfiguration mechanism.

Embedded approach for a Riemannian-based framework of analyzing 3D faces

Developing multimedia embedded applications continues to flourish. In fact, a biometric facial recognition system can be used not only on PCs abut also in embedded systems, it is a potential enhancer to meet security and surveillance needs. The analysis of facial recognition consists offoursteps: face analysis, face expressions’ recognition, missing data completion and full face recognition. This paper proposes a hardware architecture based on an adaptation approach foran algorithm which has proven good face detection and recognition in 3D space. The proposed application was tested using a co design technique based on a mixed Hardware Software architecture: the FPGA platform.

High Dynamic Range image tone mapping approach based on separable Stationary Wavelet Transform decomposition using a coefficients weighted strategy

This paper concerns the problem of converting High Dynamic Range (HDR) images into Low Dynamic Range (LDR) images so as to display the mapped images on LDR display devices. A fast, effective and flexible tone mapping algorithm preserving visibility, contrast and impression of the HDR real-world scene is developed. It is based on the Stationary Wavelet Transform (SWT) decomposition reducing a large part of halo artifacts on the reconstructed LDR image. Before the reconstruction step, the approximation and details coefficients of the coarsest subbands are weighted according to a given relationship depending on the characteristics of the HDR image. The proposed tone mapping algorithm performance is evaluated in terms of visual rendering and HDR reproduction fidelity measured with the TMQI metric. Experimental results show that the proposed algorithm provides interesting results compared to competitive methods available in the literature.