you can also train an acoustic model and use it with a voice recognition tool, like julius or sphinx. i have also just found this neural network toolkit: http://nico.nikkostrom.com/