Séance MB11 - Réseaux de neurones / Neural Network
Jour lundi, le 04 mai 2009 Salle Demers Beaulne Président Bernard K.S. Cheung
Présentations
|
15h30- 15h55 |
Numerical Solutions for Inverse Problems in Dynamic Calibration of Sensors |
|
Dan Necsulescu, University of Ottawa, Mechanical Engineering, 770 King Edward Ave, Ottawa, On, Canada, K1N 6N5 Sharareh Bayat, University of Ottawa, Mechanical Engineering, Ottawa, ON, Canada Singular Value Decomposition method can facilitate numerical computation of the generalized inverse for the resulting inverse problem and reduce the effects of random measurement noise in computer based dynamic calibration. Illustrative MATLABTM simulation results illustrate the application of Singular Value Decomposition method for dynamic calibration and permit to investigate the advantages and limitations of the proposed numerical method. |
|
15h55- 16h20 |
A Ubiquitous Positioning System that Incorporates an Indoor Wi-Fi Positioning Subsystem Using Neural Network |
|
Bernard K-S. Cheung, GERAD, École Polytechinque de Montréal, Département de mathématiques et génie industriel, Montreal, Quebec, Canada, H3C 3A7 Esmond Mok, Hong Kong Polytechnic University, Department of Land Surveying & Geo-informatics, Hong Kong, China Linyuan Xia, Sun Yat-Sen University, School of Geographical Science and Planning, Guanzhou, Guangdong, China, 510275 Ubiquitous positioning is the provision of continuous positional information in both indoor and outdoor environments. The fundamental function of Wi-Fi networks is for providing a platform for low-cost communications. We propose a possible integrated positioning system including the modeling and testing of a Neural Network based algorithm that improves the performance of the fingerprinting positioning method. |