Now that we have the first post out of the way, lets get right down to business. First, who am I? I am a member of the Sun's research lab. I am part of the speech team where we spend our time looking at issues relating to speech synthesis and speech recognition.

Most recently the speech team has been creating speech engines written entirely in the Java(TM) programming language. First, there is FreeTTS a speech synthesis engine written in Java, and there is Sphinx-4 - a speech recognition engine written in Java. Both of these engines have been released as open source under a BSD style license at SourceForge. FreeTTS has been available for a few years now, and is in use on a number of projects. We are moving toward an official Sphinx-4 release RSN.

In this blog I hope to keep track of some of the issues involved in writing on speech engines. Some issues that I'll probably touch on from time to time are:

  • Java performance
  • Improving accuracy of a recognizer
  • Open source development techniques
One of the most often questions I am asked about our work with Java Speech engines is Isn't Java too slow to meet the high performance requirements of speech? The short answer, is (of course) no. The long answer can be found in the whte paper FreeTTS - A Performance Case Study.

Post a Comment:
Comments are closed for this entry.

This blog copyright 2010 by plamere