TAPAS Training Network on Automatic Processing of PAthological Speech
Published September 13, 2018 | Version v1
Conference paper Open

Automatic Intelligibility Assessment of Parkinson's Disease with Diadochokinetic Exercises

Creators

  • 1. Universidad de Antioquia
  • 2. Ludwig-Maximilians-Universität
  • 3. Friedrich-Alexander Universität Erlangen-Nürnberg

Description

This paper presents preliminary results for the analysis of intelligibility in the speech of Parkinson’s Disease (PD) patients. An automatic speech recognition system is used to compute the word error rate (WER), the Levenshtein distance, and the similitude based dynamic time warping. The corpus of the speech recognizer is formed with speech recordings of three Diadochokinetic speech tasks: /pa-ta-ka/, /pa-ka-ta/, and /pe-ta-ka/. The data consist of 50 PD patients and 50 Healthy Controls. According to the results, the recognition error is lower for the healthy speakers (WER = 2.70%2.70%) respect to the PD patients (WER = 11.3%11.3%).

Files

wea_2018_felipe.pdf

Files (593.7 kB)

Name Size Download all
md5:2162dc2ac10d2f5fcdf9bd25a1fcdde6
593.7 kB Preview Download

Additional details

Funding

European Commission
TAPAS - Training Network on Automatic Processing of PAthological Speech 766287