• PDF / 692,107 Bytes
• 14 Pages / 439.37 x 666.142 pts Page_size
• 40 Downloads / 204 Views

DOWNLOAD

REPORT

---

Abstract. Currently, a lot of research in the field of intelligent tutoring systems is concerned with recognising student’s emotions and affects. The recognition is done by extracting features from information sources like speech, typing and mouse clicking behaviour or physiological sensors. According to the state-of-the-art support vector machines are the best performing classification models for those kinds of features. However, single classification models often do not deliver the best possible performance. Hence, we propose an approach for further improving the affect recognition performance, which is based on ideas from ensemble approaches and feature selection methods. The approach is proven by experiments on low-level speech features extracted from data which was collected in a study with German students solving mathematical tasks. In these experiments the proposed approach reached on average an affect recognition performance improvement of about 59 % in comparison to a single SVM. Keywords: Affect recognition · Low-level speech features · Intelligent tutoring systems · Feature selection · Ensemble · Classification performance improvement

1

Introduction

Automatic emotion and affect recognition is a relatively young and very important research field in the area of adaptive intelligent tutoring systems (ITSs). Some research has been done to identify useful information sources and appropriate features able to describe student’s emotions and affects. Those information sources can be speech input, written input, typing and mouse clicking behaviour or input from physiological sensors. Appropriate state-of-the-art methods for automatic affect recognition on those features are classification methods like support vector machine, k-Nearest-Neighbour, decision tree or ensembles of those (see e.g. [15,25]). The best performing state-of-the-art method in this field is, according to the literature, a support vector machine. Support vector machines are supervised machine learning methods which can be used for c Springer International Publishing Switzerland 2015  G. Conole et al. (Eds.): EC-TEL 2015, LNCS 9307, pp. 169–182, 2015. DOI: 10.1007/978-3-319-24258-3 13

170

R. Janning et al.

classification tasks and deliver in many areas the best performance in comparison to other classification approaches. Single classification models however often do not deliver the best possible performance. An improvement can be reached by means of methods like feature selection or ensemble approaches. However, ensemble approaches with the same kind of models need as input different input feature vectors and feature selection alone is not reasonable if there is already just a small amount of features. In this paper we propose an approach which overcomes this obstacles by combining ideas from feature selection and ensemble approaches for improving automatic affect recognition. The proposed approach generates one uncorrelated input feature vector per extracted feature. Subsequently, it trains classifiers (whose outputs finally are fed into a second stage class