IVML  
  about | r&d | publications | courses | people | links
   

G. Castellano, G. Caridakis, A. Camurri, K. Karpouzis, G. Volpe, S. Kollias
Body gesture and facial expression analysis for automatic affect recognition
in Scherer, K.R., Banziger, T., Roesch, E. (Eds.), A Blueprint for Affective Computing, Oxford University Press
ABSTRACT
Affect recognition plays an important role in everyday life. This explains why researchers in the human-computer and human-robot interaction community have increasingly been addressing the issue of endowing machines with affect sensitivity. Affect sensitivity refers to the ability to analyse verbal and non-verbal behavioural cues displayed by the user in order to infer the underlying communicated affect. This chapter describes affect sensitivity as an important requirement for an affectively competent agent. An affectively competent agent should be able to exploit affect sensitivity abilities to successfully interact with human users. The perception and interpretation of affective states and expressions are important to enable such an agent to act more socially and engage with users in a truly natural interaction. In this chapter, affective facial and bodily expressions are addressed as channels for the communication of affect that must be taken into account in the design of an affect recognition system. A survey of state of the art computational approaches for affect recognition, based on the automatic analysis of facial and bodily expressions and their combined information, is presented. Relevant contributions in the field are reviewed and affect sensitivity is discussed with respect to key issues that arise in the design of an automatic system for affect recognition. This chapter draws particular attention to a number of challenges in face and body affcct recognition research that have to be addressed in a more comprehensive manner in order to successfully design an affectively competent agent. First of all, an affectively competent agent should be capable of perceiving naturalistic expressions conveying states different from prototypical emotions. Second, such an agent should be able to analyse multiple modalities of expressions: new fusion methods that take into consideration the relationships and synchronization of different modalities are required. Another important issue is the dynamic account of affect and affect-related expressions: automatic analysis of temporal dynamics is required for an affective competent agent to be able to monitor the evolution of the affective states displayed by the user over time. Robustness in the real-world environment is an issue not to be neglected as, to date, many automatic systems for the analysis of affective facial and bodily expressions only work well in controlled environments. Finally, context sensitivity is currently an undcrexplorcd topic that has to be taken into consideration in the design of an affectively competent agent, as the detection of the most subtle affective states can be achieved only through a comprehensive analysis of their causes and effects.
27 July , 2010
G. Castellano, G. Caridakis, A. Camurri, K. Karpouzis, G. Volpe, S. Kollias, "Body gesture and facial expression analysis for automatic affect recognition", in Scherer, K.R., Banziger, T., Roesch, E. (Eds.), A Blueprint for Affective Computing, Oxford University Press
[ save PDF] [ BibTex] [ Print] [ Back]

© 00 The Image, Video and Multimedia Systems Laboratory - v1.12