In the end, the language configuration will be based on text files, so any user could personalize their Magnus so that it recognized any command they wanted. So if anyone needs to use Magnus in Spanish, for example, it will be just a matter of editing a few text files with some phonetic transcriptions, and the process will be over and ready to run smoothly. In the documentation that will be available near the end of (this first stage of) the project, say my Master Thesis, all the aspects of functionality and configuration will be thoroughly explained.

Let me thank you very much for your interest in Magnus. I invite you to keep in touch with it proposing interesting lines of development for Guadalinex. I hope that in a few day’s time, once I’m over with my exams, a first release of Magnus (0.1.0) will be available for testing, along with a list of improvements, your suggestion being among them, of course.

On the other side Magnus uses a limited vocabulary, which enables a better accuracy to be obtained since it can focus on details, say, different phonetic transcriptions for a single command. This task would be very difficult if thousands of words were contemplated with this system.

So I believe it’s difficult to say when the goal of your question, Juaxix, will be accomplished.

I’m really interested in it for Guadalinex !

As far as I am concerned, statistical analysis of speech (Hidden Markov Models) is more accurate than deterministic analysis (Patterns Recognition) when it comes to automatic speech recognition. That’s the main fact Sphinx-4, based on HMMs, was chosen to be the base of magnus instead of alternatives based on Patterns Recognition such as Vocal Joystick in this case.

It can be seen in the VJ Video Demonstrations that the various systems are mainly controlled with the voice’s pitch. The goal of magnus, on the other side, is to provide a means of mouse control via several voice keywords. It’s a different way of dealing with the problem. Not better nor worse, just different. As usual in engineering development, it all depends on the the final application.

Another interesting project based on Patterns Recognition applied to people with motor impairments and also carried out in the Ministry of Education can be found in Jordi Lagares’s Fressa Project.

Echale un vistazo.