Simon 0.4 begins listening
The Simon open source speech recognition system has been updated with a new recognition layer and context-awareness to improve accuracy and performance. The Simon project, which sets out to be a replacement for a mouse and keyboard, especially for people with physical disabilities, moved under KDE's umbrella in April 2012, becoming part of KDE Extragear.
In Simon 0.4, work on making it easier to use has resulted in a reorganised main window with all the important options on one screen. The addition of a Simon base model and integration of a repository of base models means that it is quicker to setup; the developers say it is possible to go from a fresh install to a working system in 5 minutes with less user interaction than before. A video steps through the install process:
Simon 0.4 internally supports the BSD-licensed CMU SPHINX toolkit for speech recognition and has an English VoxForge model available. Training samples gathered by Simon can now be contributed to the VoxForge project to help train its development of general GPL acoustic models for speech recognition.
Context enhancements mean that Simon 0.4 uses "context conditions" to help it decide what it is looking for and to reconfigure context on the fly. Among the techniques used, Simon is able to use the webcam to watch the user and look for lip movement or, more traditionally, it can detect what windows are currently open and what has focus and use that information to decide on which scenarios it should be listening for. Another video shows this in action:
Other enhancements include the new scripted dialog system which aids the creation of specific speech-based interaction, a KDE plasmoid for monitoring Simon, a new sample review tool, and an Akonadi command plugin. Simon 0.4.0 is available, under a GPL licence with some caveats, as a Windows installer or as source code.