Oudeyer, P-Y. The production and recognition of emotions in speech: features and algorithms. International Journal of Human Computer Interaction, 59(1-2):157-183, 2003. special issue on Affective Computing

Sony CSL authors: Pierre-Yves Oudeyer

Abstract

Recent years have been marked by the development of robotic pets or partners such as small animals or humanoids. The interactions with them are very different from those with traditional computers : instead of having human beings using robotic conventions, robots should learn to communicate in a humanised fashion. In par- ticular, they need to be able to express and recognize emotions. This can be done in part using speech, which has the advantage to be computationally cheap and practical to implement in real world robots. Nevertheless, research in this area is still very young. We present here algorithms that allow a young robot to express its emotions like babies do. They are very simple and eƆciently provide life-like speech thanks to the use of concatenative speech synthesis. We describe a technique which allows to control continuously both the age of a synthetic voice and the quantity of emotions that are expressed. This is useful since personal robots may grow up and have many degrees of emotions. Also, we present the first large-scale data mining experiment about the automatic recognition basic emotions in unformal everyday short utterances. We focus on the speaker dependant problem. We compare a large set of machine learning algorithms, ranging from neural networks, Support Vector Machines or decision trees, together with 200 features, using a large database of several thousands examples. We show that the difference of performance among learning schemes can be substantial, and that some features which were previously unexplored are of crucial importance. An optimal feature set is derived through the use of a genetic algorithm. Finally, we explain how this study can be applied to real world situations in which possibly very few examples are available. Furthermore, we describe a game to play with a personal robot which allows to teach it examples of emotional utterances in a natural and rather unconstrained manner.

Keywords: emotion, speech, robots, production, recognition

Downloads

[PDF] Adobe Acrobat PDF file

BibTeX entry

@ARTICLE { oudeyerijhcs, AUTHOR="Oudeyer, P-Y.", JOURNAL="International Journal of Human Computer Interaction", NOTE="special issue on Affective Computing", NUMBER="1-2", PAGES="157--183", TITLE="The production and recognition of emotions in speech: features and algorithms", VOLUME="59", YEAR="2003", }