state of speech recognition and speech driven handicap accessibility on Linux
People who use speech recognition software usually want it to do two things: capture text at the speed of sound, and navigate the keyboard without touching the keys. Until now, both types of speech recognition have been available only on Windows, and the Windows market is dominated by one product: Dragon NaturallySpeaking. Very recently, and this has changed. Speech recognition of both types is now within the Linux-user's grasp, thanks to two entirely different programs with different approaches to solving the Linux speech-recognition problem.
For those who want to enter a lot of text quickly such as this article,, our good friends at Wine have enabled Dragon NaturallySpeaking 8 to run on Linux through Wine--and run it runs VERY well. NaturallySpeaking transcribes my speech as quickly as I can speak as long as I confine my dictation to the included DragonPad wordprocessor. NaturallySpeaking cannot communicate or interoperate with any other application outside of the Wine environment, so cut-and-past is nearly useless.
for users that are content with a simple grammar for command and control for the desktop, Perlbox Voicehttp://www.perlbox.orgwill give you control of your keystrokes. Installation of both programs was easy. Pearlbox Voice is available on Sourceforge in rpm, deb, or tarball. Installation of DNS8 is easy but not intuitive, and the directions may change slightly as Wine is developed. For a look at the current instructions, see the WineHQ site or http://ossri.harvee.org.
While this development is exciting for the state of handicap accessibility for those of us with damaged or otherwise dysfunctional hands, it is not sufficient. We need help, people with working hands to help write code.there are a number of partially completed projects out there for people with upper extremity disabilities.
- voice coder, translates a limited set of spoken English into Python. Comes complete with ability to change the style of variable naming (joined by underscores, Bumpy Caps etc.)
- vr-mode: enables select and say for Emacs, both local and remote. Very useful when using virtual machines.
natlink: enabling technology for the above. Links NaturallySpeaking to Python.
All three of these projects need help. The tasks range from increasing stability to improving ease of install and the ever popular problem in speech recognition. "is it a bug, or is it misrecognition?"
Recognizing that solving these problems is not easy and requires a legal structure, we have formed the "Open Source Speech Recognition Initiative" a 503C nonprofit. We are looking for people who will help volunteer some time to make it possible for speech recognition dependent disabled people to become employable and self-sufficient on Linux.
We look forward to your help.