Bertrand Ravera

Learn More
The CARETAKER 1 project, which is a 30-month project that has just kicked off, aims at studying, developing and assessing multimedia knowledge-based content analysis, knowledge extraction components, and metadata management subsystems in the context of automated situation awareness, diagnosis and decision support. More precisely, CARETAKER will focus on the(More)
This paper proposes an unsupervised method for real time detection of abnormal events in the context of audio surveillance. Based on training a One-Class Support Vector Machine (OC-SVM) to model the distribution of the normality (ambience), we propose to construct sets of decision functions. This modification allows controlling the trade-off between(More)
This paper describes a new very low bit rate speech coder at 600 bits/s based on the MELPe algorithm (A.V. McCree and T.P. Barnwell III, 1995). This coder associated with the other bit rates at 2400 bits/s and 1200 bits/s constitute the new NATO standard STANAG 4591. The STANAG 4591 noise pre-processor is also part of the 600 bits/s coder. At 600 bits/s,(More)
In this demonstration, we will show the different modules related to the automatic surveillance prototype developed in the context of the EU VANAHEIM project. Several components will be demonstrated on real data from the Torino metro. First, different unsupervised activity modeling algorithms that capture recurrent activities from long recordings will be(More)
Even if the speaker recognition field is very dynamic, few studies concern the constraints linked to the use of a speaker recognition system inside a professional telecommunication network. This paper deals with this problem and proposes some adaptation of such system in the focus of a real world network monitoring application. This work is specifically(More)
– Cet article introduit une méthode non-supervisée pour la détection en temps réel d'événements sonores anormaux dans le contexte de la surveillance audio. En se basant sur les machines à vecteur de support 1-classe (OC-SVM) pour modéliser une distribution, celle d'une ambiance sonore, nous introduisons une hypothèse complémentaire et proposons de(More)
– Nous introduisons un problème SVM 1-classe sans biais, exploitant les contraintes du problème 2-classes. A travers l'écriture d'une forme duale unifiée des problèmes SVM 1-classe et 2-classes, nous montrons qu'il est possible d'utiliser l'algorithme Sequential Maximization Gradient Optimization (SMGO), indépendemment du problème. Ce résultat fait(More)
This paper presents a multi-modal health and activity monitoring framework that enables abnormal event detection and long term evaluation of the health of elderly people at home. We describe its integration with the ESTIA [1] Residential Gateway and detail the demonstration scenario we developed involving different kinds of sensing modules.