Project

Back to overview

Behavioural Modelling of Human Experts for Scene Analysis

English title Behavioural Modelling of Human Experts for Scene Analysis
Applicant Bierlaire Michel
Number 117823
Funding scheme Project funding (Div. I-III)
Research institution Laboratoire transport et mobilité EPFL - ENAC - INTER - TRANSP-OR
Institution of higher education EPF Lausanne - EPFL
Main discipline Mathematics
Start/End 01.10.2007 - 30.09.2010
Approved amount 263'100.00
Show all

Keywords (8)

Behavioral model; image analysis; Discrete Choice Analysis; Scene Analysis; Pattern recognition; Computer Vision; Machine Learning; Human-Machine Interfaces

Lay Summary (English)

Lead
Lay summary
In this project we are interested in modelling the global appraisal process of experts in the context of scene analysis. The idea is to develop a methodological framework around a behavioural model capturing that process. This behavioural model must be able to capture the observable causal actions of the decision process of the expert, but also the intrinsic uncertainty which must lead to a heterogeneity of decisions across experts. What we propose is an original approach based on discrete choice models, that are econometric models designed to predict the behaviour of decision-makers faced to a choice.

The developed model will be designed to predict the appraisal of a human expert in charge of labelling a video sequence representing a facial expression, taking into account objective information (including features that can be automatically extracted from the video feed) and potential expert-speci?c subjective judgement. The main source of data will be collected using a web-based survey where various individuals will have to label several facial expressions appearing in video sequences. A discrete choice model based on (random) utility maximisation principle will be estimated and validated.

Although we are mainly interested in fundamental research and methodological aspects, we will consider how the model can be used for real applications, such as the automatic analysis of facial expressions of drivers, the meeting supervision in smart room scenarios or the intelligent human-machine interfaces.
Direct link to Lay Summary Last update: 21.02.2013

Responsible applicant and co-applicants

Employees

Associated projects

Number Title Start Funding scheme
141099 Pedestrian dynamics: flows and behavior 01.04.2012 Project funding (Div. I-III)

-