Automatic Speech Recognition for Electro-Laynx Speech

Seminar Type: Speech Processing, Seminar (3 SE)
Student: Gabriel Hülser, Felix Rothmund

table.png

 Short Description

The electro-larynx is a small, hand-held and battery driven device which is used to re-obtain speech when the larynx of a person is removed. The resulting speech will sound very unnatural and the performance in Automatic Speech Recognition (ASR) systems will significantly decrease which is indicated in the table.

ASR system are increasingly part of daily life and therefore, the word accuracy rate of electro-larynx speech should be reasonably high in order to be able to make ASR technologies accessible for patients suffering from speech disorders.
In previous work electro-larynx speech is applied on an ASR system which was designed for normal, healthy speech [1]. 
In this project different types of features for the ASR should be tested. Furthermore the relationship between ASR performance and intelligibility should be investigated.
 

Your Tasks

  • Review of literature
  • Testing different features and scenarios for electro-larynx speech in ASR
  • Documentation

Your Profile

  • Motivation and interest in the topic
  • Background information in  HTK, Matlab and (speech) signal processing

Full text

The final text of this thesis can be found here.

Contact

Martin Hagmueller, Anna Katharina Fuchs ({hagmueller, anna.fuchs}@tugraz.at or 0316/873 4377; 0316/873 4367)

References

[1] A.K. Fuchs, J.A. Morales-Cordovilla and M. Hagmueller, "ASR for Electro-Laryngeal Speech". in Proc. ASRU, Olomouc, Czech Republic, 2013. 

 

AttachmentSize
ASRforELSpeech_HuelserRothmund_Bachelor-thesis.pdf534.68 KB