WIT Press


Fear-type Emotion Recognition And Abnormal Events Detection For An Audio-based Surveillance System

Price

Free (open access)

Paper DOI

10.2495/RISK080461

Volume

39

Pages

9

Page Range

471 - 479

Published

2008

Size

1,673 kb

Author(s)

C. Clavel & T. Ehrette

Abstract

Fear-type emotion recognition and abnormal events detection for an audio-based surveillance system C. Clavel & T. Ehrette Thales Research and Technology France, RD 128, 91767 Palaiseau Cedex, France Abstract The goal of our research work is to carry out an audio-based abnormal situation diagnosis system for the SERKET project which aims at developing surveillance systems dealing with dispersed data coming from heterogeneous sensors.We look at things from the point of view of human life protection in the context of civil safety. Therefore, we focus on abnormal situations during which human life is threatened (psychological and physical attack). The proposed system relies on information conveyed by both speech and non-speech acoustic manifestations to generate alerts. More precisely, our audio module can be divided into two sub-modules. The first one concerns the abnormal event detection system that is illustrated here in the case of gun-shot. The second one focuses on information conveyed by the emotional content of speech. Such information is useful for decoding human behaviour in abnormal situations and so provides a situation diagnosis. More precisely the targeted emotions are fear-type emotions corresponding to symptomatic emotions occurringwhen the matter of survival is raised, including the different fear-related emotional states from worry to panic. At the last stage of its development, this work would propose a surveillance system plugged into a real control room. Thus, we proposed here a mock-up to illustrate the running of these two systems. 1 Introduction Civil safety has received growing interest over the last few years. Recent projects such as SERKET (http://www.research.thalesgroup.com/software/cognitive solutions/Serket/index.html) tackle the issue of the development of automatic

Keywords