Second Workshop on Child, Computer and Interaction (WOCCI 2009)

Cambridge, MA, USA
November 5, 2009

Automatic Classification of Reading Disorders in a Single Word Reading Test

Andreas Maier (1), Stefanie Horndasch (2), Elmar Nöth (3)

(1) Abteilung für Phoniatrie und Pädaudiologie, Universitätsklinikum Erlangen, Gwermany
(2) Kinder- und Jugendabteilung für Psychische Gesundheit in der Psychiatrischen und Psychotherapeutischen Klinik, Erlangen, Germany
(3) Lehrstuhl für Mustererkennung, Universität Erlangen-Nürnberg, Erlangen, Germany

In clinical practice, reading disorders are still evaluated perceptually. In order to alleviate this problem, we propose to use automatic speech processing techniques to classify reading disorders. Therefore, we recorded 38 children who were suspected to have a reading disorder. The recordings were performed using a German standard test for reading disorders. Each child was perceptually assessed and the number of reading errors per child was recorded. Furthermore, the reading duration was stored for each child. If either of both values exceeded an age-dependent limit, the child was diagnosed having a reading disorder. In 30 of the 38 children the reading disorder was confirmed. In this paper, we present the results on the automatic evaluation concerning a single word reading test. We achieve up to 78.9% recognition rate in the detection of the exceedance of the reading error limit and 97.4% recognition rate in the classification for reading disorder.

Full Paper

Bibliographic reference.  Maier, Andreas / Horndasch, Stefanie / Nöth, Elmar (2009): "Automatic classification of reading disorders in a single word reading test", In WOCCI-2009, 27-30.