Evaluating synthetic speech

Title:
Review of methods for evaluating synthetic speech.

Author:
Heather Cryer, Sarah Home.

Publisher:
RNIB Centre for Accessible Information.

Year of publication:
2010.

Background

This paper highlights the different approaches to synthetic voice evaluation, exploring the different purposes of evaluation, advantages and disadvantages of particular methods and the complexities involved.

Key findings

  • Objective/acoustic measures are largely used by developers of synthetic voices to identify where problem areas lie
  • User testing is carried out by people developing products or services with synthetic speech, to see whether users like the voice and how well they understand it
  • Feature comparisons are used by people trying to choose the right voice for a particular application, to see whether it has the functionality they require

Downloads

Last updated: 20 February 2013

Make a donation

Right now we can only reach one in three of the people who need our help most.

Please make a donation and help us support more blind and partially sighted people.