Length analysis of speech to be recorded in the recognition of Parkinson's disease

Jenei Attila Zoltán and Sztahó Dávid: Length analysis of speech to be recorded in the recognition of Parkinson's disease. In: Magyar Számítógépes Nyelvészeti Konferencia, (18). pp. 137-149. (2022)

[thumbnail of msznykonf_018_137-149.pdf] Cikk, tanulmány, mű
msznykonf_018_137-149.pdf

Download (379kB)

Abstract

Parkinson's disease is an incurable neurodegenerative disease to the present clinical knowledge. It is diagnosed mostly by exclusion tests. Numerous studies have confirmed that speech can be promising to suspect the presence of the disease. On the other hand, just a few researches discuss the appropriate length of the speech sample or the contribution of parts of the full-length recordings in the classification. Hence, we partitioned each original recording into four shorter samples. We trained linear and radial basis function (rbf) kernel Support Vector Machine (SVM) models separately for original recordings, each partitioned group and all partitioned samples together. We found no significant difference between the results of the rbf kernel models. However, we obtained significantly better results with a portion of the entire speech using linear kernel models. In conclusion, even a shorter piece of a longer speech may be adequate for classification.

Item Type: Article
Heading title: Beszédtechnológia
Journal or Publication Title: Magyar Számítógépes Nyelvészeti Konferencia
Date: 2022
Volume: 18
ISBN: 978-963-306-848-9
Page Range: pp. 137-149
Language: English
Place of Publication: Szeged
Event Title: Magyar számítógépes nyelvészeti konferencia (18.) (2022) (Szeged)
Related URLs: http://acta.bibl.u-szeged.hu/75797/
Uncontrolled Keywords: Nyelvészet - számítógép alkalmazása, Beszédtechnológia, Parkinson-kór
Additional Information: Bibliogr.: p. 147-149. ; összefoglalás angol nyelven
Subjects: 01. Natural sciences
01. Natural sciences > 01.02. Computer and information sciences
06. Humanities
06. Humanities > 06.02. Languages and Literature
Date Deposited: 2022. May. 24. 15:56
Last Modified: 2022. May. 24. 15:56
URI: http://acta.bibl.u-szeged.hu/id/eprint/75870

Actions (login required)

View Item View Item