Showing papers by "Grenoble Institute of Technology published in 1999"

Real-time DSP implementation for MRF-based video motion detection

Abstract: This paper describes the real time implementation of a simple and robust motion detection algorithm based on Markov random field (MRF) modeling, MRF-based algorithms often require a significant amount of computations. The intrinsic parallel property of MRF modeling has led most of implementations toward parallel machines and neural networks, but none of these approaches offers an efficient solution for real-world (i.e., industrial) applications. Here, an alternative implementation for the problem at hand is presented yielding a complete, efficient and autonomous real-time system for motion detection. This system is based on a hybrid architecture, associating pipeline modules with one asynchronous module to perform the whole process, from video acquisition to moving object masks visualization. A board prototype is presented and a processing rate of 15 images/s is achieved, showing the validity of the approach.

An unknown input observer design for linear time-delay systems

Abstract: This paper presents an algebraic method to design a dynamic gain unknown input observer for linear systems with commensurate delays in state, input and output variables. Necessary and sufficient conditions for the existence of such an observer are provided as well as a constructive procedure. An example is given in order to illustrate the proposed method.

A new SNR-feature mapping for robust multistream speech recognition

Abstract: Keywords: glotin ; speech Reference EPFL-CONF-82548 Record created on 2006-03-10, modified on 2017-05-10

A CASA-labelling model using the localisation cue for robust cocktail-party speech recognition

Abstract: Keywords: glotin ; speech Reference EPFL-CONF-82547 Record created on 2006-03-10, modified on 2017-05-10

Causal modeling for supervision

Abstract: Today, process control and monitoring are evolving to include plant safety and availability management, online diagnosis and maintenance policy. Human operators are at the highest hierarchical level in the organization of the control system. Supervision aims are to assist control operators in their decision-making tasks, to help them understand and identify operating situations underway in a man-machine cooperation objective. The paper demonstrates how developments in causal reasoning can contribute judiciously to the supervision problem. An original causal modeling that includes time management is described and used for simulation, explanation and diagnosis.

Generation of defects with diamond and silica particles inside high-reflection coatings: influence on the laser damage threshold

Abstract: High power lasers at 1.06 micrometers wavelength like the French 'Megajoule' program require large area mirrors made of alternative layers of HfO2 and SiO2 with high damage threshold. It has ben already identified that nodular defects have a significant role in the activation of laser damage. The purpose of this paper is to study experimentally the role played by the size and the nature of the seed on the critical behavior of the nodular defects under laser irradiation. Artificial nodules have been grown with careful control of location and seed size. these nodules have been generated with diamond and silica seeds. Silica seeds of four different sizes have been used. A method which minimizes clusters of particles has been successfully used. With this method, we have obtained defects with a sharp size distribution. These samples have been tested on our automatic damage test bench in various modes. The thresholds between polluted and unpolluted mirrors are discussed according to the size and the nature of the seeds. The damage sites are observed with a scanning electron microscope. These experimental results are compared with a damage model.© (1999) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

A measure of speech and pitch reliability from voicing

Abstract: Keywords: glotin ; speech Reference EPFL-CONF-82549 Record created on 2006-03-10, modified on 2017-05-10

Iterative improvement based multi-way netlist partitioning for FPGAs

Abstract: This paper presents a multi-way FPGA partitioning method. The basic idea is similar to one proposed by Kuznar et al. (1995), but instead of using the replication and re-optimization, it takes force of the classical iterative improvement partitioning techniques. The basic effort consists in guiding the classical algorithms in their solution space exploration. This was done by introducing the cost function based on the infeasibility distance of the partitioning solution and carefully tuning the basic parameters of the classical algorithms such as definition of size constraints for feasible moves, handling solutions stack, selecting best cluster to move, etc. The proposed method obtains results comparable to the best published results, and even outperforms them for the largest benchmarks.

2-Dimensional stochastic resonance in a discrete time nonlinear system with a noisy elliptic input.

Abstract: This paper deals with stochastic resonance. This nonlinear physical phenomenon generally occurs in bistable systems excited by a random noise plus a sine. Such systems force cooperation between the input noise and the input sine: Provided a fine tuning between the power noise and the dynamics, the system reacts periodically. The interesting fact is that the local output signal-to-noise ratio presents a maximum when plotted against the input power noise. In this paper we recall the results for the discrete-time one-dimensional nonlinear AR(1) systems. We then extend the study to a particular 2-dimensional nonlinear system.