Fixed, small vocuabulary, language independent speech (or command) recognition

Question

Looking for any low-computational-complexity recognition engine (and hopefully FOSS), that works well for:

1. fixed-set / small-vocabulary (upto 100 phrases)
2. language-independent
3. discontinuous speech / commands

It is fine, if the engine is:

1. speaker-dependent
2. (thus) requires a-priori training (before recognition can be performed)

By low-computational-complexity, I mean something which I can hope to run on a 450MHz ARM9 SBC with 64MB RAM.

Answer 1

Texas Instruments has an open source recognition engine for embedded platforms. Is it for small and medium vocabularies, however you might have to make the models yourself if you need to support many languages.

https://gforge.ti.com/gf/project/tiesr/