Child affective speech recording and its acoustic analysis in a cross-linguistic perspective

I. Grichkovtsova and I. Mennen

Queen Margaret University College, Edinburgh, UK

This paper presents the developed methodology for the cross-linguistic study of the production of child affective speech. Affective information plays a very important role in speech and communication. Recent cross-linguistic studies show that there is a number of available means (pitch range, intensity, voice quality, intonation contours, rhythm, etc) to express affect in speech, but their usage, level of importance and meaning vary in different languages. Affective speech research is mainly concerned with adults, and studies of child affective speech are very rare and mostly deal with perception only. The main reason for this is the methodological difficulties of adult, and especially child affective speech recording. After providing a short summary of these methods (spontaneous, acted and induced affective speech) and their discussion in the light of child affective speech recording, the chosen methodology for this study will be described.

For ethical reasons, only acting and light induction methods were chosen to record child speech. Visual materials were developed in a realistic manner together with a professional illustrator and based on the research of facial expressions. They represent a child expressing four emotions (happiness, sadness, anger and fear); the child gender and age are matched to that of recorded children. The randomized cards have 16 repetitions for each emotion. The subject says one token utterance in the same way as the child at the picture, thus the child is playing emotions through association with the drawing. One token utterance was selected with a similar sound and prosodic structure for English and French. 16 monolingual and 8 bilingual children were recorded in total. Bilingual children were recorded in two sessions (one for each language) by different speakers (to avoid code-switching); there was a period of about two weeks between the sessions.

This methodology enabled us to record a cross-linguistically comparable corpus of bilingual Scottish-French children and their monolingual counterparts. Acoustic measurements of pitch range variation, intonation contour, timing of intonation patterns and; voice quality are taken. The results compare the emotions across languages and across children.

Paper presented at Measuring Behavior 2005 , 5^th International Conference on Methods and Techniques in Behavioral Research, 30 August - 2 September 2005, Wageningen, The Netherlands.

© 2005 Noldus Information Technology bv