13th Annual Conference of the International Speech Communication Association

Portland, OR, USA
September 9-13, 2012

Synthetic Correction of Deviant Speech – Children's Perception of Phonologically Modified Recordings of Their Own Speech

Sofia Strömbergsson

Department of Speech, Music and Hearing, School of Computer Science and Communication, KTH (Royal Institute of Technology), Stockholm, Sweden

This report describes preliminary data from a study of how children with phonological impairment (PI) perceive automatically corrected versions of their own deviant speech. The results from 8 children with PI are compared to results of a group of 20 children with typical speech and language (nPI). The results indicate group differences only in tasks where the children make judgments of their own recorded (original or modified) speech; here, the children in the nPI group perform significantly better than the children with PI. In tasks where the children judge the phonological accuracy of recordings of other children (original or modified), however, the two groups perform equally well. Furthermore, the results indicate that sub-phonemic modifications of recordings are too subtle for the children in both groups to detect. Technical and clinical implications of these findings are discussed.

Index Terms: speech perception, children, phonological impairment, unit selection

Full Paper

Bibliographic reference.  Strömbergsson, Sofia (2012): "Synthetic correction of deviant speech – children's perception of phonologically modified recordings of their own speech", In INTERSPEECH-2012, 1131-1134.