Automatic Recognition of Non-Acted Affective Postures

Abstract

The conveyance and recognition of affect and emotion partially determine how people interact with others and how they carry out and perform in their day-to-day activities. Hence, it is becoming necessary to endow technology with the ability to recognize users' affective states to increase the technologies' effectiveness. This paper makes three contributions to this research area. First, we demonstrate recognition models that automatically recognize affective states and affective dimensions from non-acted body postures instead of acted postures. The scenario selected for the training and testing of the automatic recognition models is a body-movement-based video game. Second, when attributing affective labels and dimension levels to the postures represented as faceless avatars, the level of agreement for observers was above chance level. Finally, with the use of the labels and affective dimension levels assigned by the observers as ground truth and the observers' level of agreement as base rate, automatic recognition models grounded on low-level posture descriptions were built and tested for their ability to generalize to new observers and postures using random repeated subsampling validation. The automatic recognition models achieve recognition percentages comparable to the human base rates as hypothesized.

Naturalistic Affective Expression Classification by a Multi-stage Approach Based on Hidden Markov Models

“…In the affective computing field [12], various studies have been carried out to create systems that can recognize the affective states of their user by analyzing their vocal [1], facial [11] [17], and body expressions [4], and even their physiological changes [6]. Most of the work has been carried out on acted or stereotypical expressions.…”

Automatic Recognition of Affective Body Movement in a Video Game Scenario

“…To evaluate the performance of the automatic recognition system, we followed a simplified version of the method proposed in [19]. The evaluation method proposed in [19] requires three groups of observers in order to fully separate the computation of the benchmark from the testing of the system.…”

“…Only 36% accuracy was, instead, obtained for 'concentration'. The low accuracy obtained for 'concentration' could be due to the fact that the human observers may have used this label when the avatar's expression did not express any of the other affective states as discussed in [19]. Finally, the column named as 'Multiple classes' in table 3 contains the number of the test samples that our algorithm was not able to categorize into only one class.…”

“…A study that aims at detecting emotional states from non-acted body expression is presented in [19]. This is very relevant to our work as the scenario considered is whole-body computer games.…”

“…We then used the most frequent label associated by the observers to an avatar as the representative affective state for that avatar. We call this label ground truth following the approach used in [19]. Fig.…”