Hi,
When Jelly Bean was released, one of its features was offline voice dictation through its built-in speech recognition engine (SREC by Nuance Software). Its release in Jelly Bean effectively means that it's under the Apache License.
I've downloaded the Android source and extracted the speech recognition engine as well as the acoustic model (US English) to my github repository https://github.com/arthurv/srec.
I've been meaning to port it into Debian (Raspbian) however I've noticed that it uses the Java interface (JNI) to interface with the microphone as well as the apps that use it, and being unfamiliar with the Android system architecture, I was wondering if there's anyone in these forums who can help in porting this system to plain linux? Documentation is available from my github repo, I've been reading it for a while but I don't have the time nor Android dev experience to accomplish this. If someone with Android knowledge gets involved, I will do all I can to help out.