UCLIC Affective Body Posture and Motion Database
Description:

The UCLIC Affective Body Posture and Motion Database consists of affective motion capture recordings in 2 different settings.

Naturalistic body motions of 11 people in a video game context (playing Wii sports). This data was collected with a Gypsy5 electro-mechanical motion capture system (Animazoo UK Ltd.) http://www.metamotion.com/gypsy/gypsy-motion-capture-system.htm.

Acted body motions of 13 non-professional actors portraying angry, fear, happy and sad emotions. This data was collected with a Vicon MX series digital optical motion capture system http://www.vicon.com/.

Access:

The database is free for research and educational purposes.
If you already have a username and password, login here.

Otherwise, please email n.berthouze@ucl.ac.uk or alk@cise.ufl.edu> with the following information:

  • Your name
  • Your research institute (name, city, country, web address if available)
  • Some words on your research and how the database would be used in it
  • How you heard of the database (colleague, Internet, etc.)

We will then send you a username and password for accessing the database.
Papers:

Acted corpus:
A. Kleinsmith, R. De Silva, N. Bianchi-Berthouze, "Cross-Cultural Differences in Recognizing Affect from Body Posture", Interacting with Computers, 18 (6), 1371-1389, 2006

N. Bianchi-Berthouze, A. Kleinsmith "A categorical approach to affective gesture recognition", Connection Science, V. 15, N. 4, pp. 259-269, 2003

Andrea Kleinsmith, Nadia Bianchi-Berthouze, "Recognizing Affective Dimensions from Body Posture", Proceedings of the Int. Conf. of Affective Computing and Intelligent Interaction, LNCS 4738, 48-58, Lisboa (Portugal), September 2007

Naturalistic corpus:
Kleinsmith, A., Bianchi-berthouze, N., Steed, A. (2011). Automatic Recognition of Non-Acted Affective Postures. IEEE Transactions on Systems, Man and Cybernetics, Part B , 1-12

Citations:

If you use the acted corpus , please cite the following paper:
A. Kleinsmith, R. De Silva, N. Bianchi-Berthouze, "Cross-Cultural Differences in Recognizing Affect from Body Posture", Interacting with Computers, 18 (6), 1371-1389, 2006

If you use the naturalistic corpus, please cite the following paper:
Kleinsmith, A., Bianchi-berthouze, N., Steed, A. (2011). Automatic Recognition of Non-Acted Affective Postures. IEEE Transactions on Systems, Man and Cybernetics, Part B , 1-12

You may also be interested in the survey paper: Kleinsmith, A., Bianchi-Berthouze, N. (2013). Affective Body Expression Perception and Recognition: A Survey. IEEE Transactions on Affective Computing 4(1), 15-33
Ethical Clearance Issues:

The motion capture data consists of the numerical positions of body joints in a 3D space, therefore the anonymity of the participants is preserved.

Ethical clearance has been obtained for the naturalistic data.

The acted data was collected at a university which does not require ethical clearance to be obtained.

Purpose of collection:

To create an automatic affective recognition system using body posture.
Format:

Naturalistic: .bvh files

Acted: .csm files
Elicitation method:

Naturalistic: Non-acted affective body postures and motions were exhibited by participants while playing Wii sports games. The participants were not aware of the purpose of the study.

Acted: Participants were asked to perform their own idea of angry, fear, happy and sad emotions through bodily expression. No constraints were placed on the actors in how they performed each emotional posture.
Size:

Naturalistic: 36 motion capture files and 105 static posture images in .jpg format.

Acted: 183 motion capture files and static posture images in .jpg format.
How was emotion determined?:

Naturalistic: Labelling is still underway.

Acted:

* Emotion category:

  • 112 postures have been labelled by 3 different groups of observers (25 Japanese, 25 Sri Lankans and 20 Caucasian Americans) according to 8 emotion labels (angry, depressed, fear, joyful, happy, sad, surprised and upset).
  • The entire set of 183 postures are also labelled according to the actors' intended emotion (angry, fear, happy and sad).

* Affective dimensions:

  • 112 postures have been labelled by a group of 10 observers according to 4 dimensions arousal, valence, potency and avoidance. A 7-point Likert scale was used. The labels are still being evaluated and are not yet available with the database. When the labels have been fixed, an email will be sent to those interested.