Speech Therapy Project
Student: James Richards (MEng final year)
Supervisor: Dr P.J.Leonard
Project aims
The goal of the project
was to develop a system that provides visual feedback on a users voice.
One use of the system could be in speech therapy as a additional
learning tool. Users will learn from clear visual feedback how to
adjust their vocal tract and lips to more closely imitate the ideal
vowel sound.
How it works
The system uses a voice recognition algorithm that has been
'trained' on a number of peoples voices. The computer learns the
characteristics of the sound presented (using a neural net). Once trained it can be
used to provide a comparison against any new input voices.
Training Set
Training data was obtained by recording 12 different people saying the
following phonemes:
AHH father
EEE
flee
EHH bed
UHH burn
ERR grrreat
OOH
blue
Visual Feedback
The program classifies the sound the user is making and provides instant visual feedback via on-screen graphics.
A simplified diagram of the vocal tract is morphed in
real-time.
Demo
Because the project software was written in JAVA so it is possible
to run the program from a browser. You will need to have a
microphone in order to use the
program. Some cheap microphones can be quite noisy. For this application the Logitech USB
Desktop Microphone seems to work quite
well (about 15 pounds from amazon) . The applet runs as a group of floating windows.
Note the program will exit if you don't have a micrphone connected!
CLICK HERE FOR DEMO
Comments or questions? Please contact p.j.leonard@bath.ac.uk
More music / speech projects.