Signal Processing and Speech Communication Laboratory
hometheses & projects › EL Speech improvement on a Raspberry Pi

EL Speech improvement on a Raspberry Pi

Status
Open
Type
Master Project
Announcement date
05 May 2015
Mentors
Research Areas

Short Description

The larynx, which contains the vocal folds, is essential for the production of speech. If the larynx is surgically removed, typically due to cancer, the patient loses the ability to speak. One possibility to be able to communicate again is to use a hand-held, battery-driven device – the electro-larynx. The device is held against the neck and its vibrations are transmitted through the neck. Then the patient can create speech by modulating the energy in the vocal tract. One drawback of the resulting speech is that it sounds robotic.

In previous work algorithms have been developed to enhance the electro-larynx speech in terms of naturalness and intelligibility. So for the programs run on a personal desktop. In order to make the system accessible for many applications an implementation on an embedded computer would be desirable.  The Raspberry Pi is such a low-cost embedded computer, where we would like to run existing enhancement strategies to make the whole system portable.

Your Tasks

  • Installation of CMU pocketsphinx on a Raspberry Pi
  • Implementation of existing algorithms
  • Documentation

Your Profile

  • Motivation and interest in the topic
  • Knowledge in Matlab, Phyton, Linux and (speech) signal processing

Contact

Anna Katharina Fuchs (anna.fuchs@tugraz.at or 0316/873 4367)