Babel released the PocketBabEar version 2.0 speech recognition system 

  PocketBabEar is a full-featured Speech Recognition software product designed for the 
  embedded market. PocketBabEar has been designed to minimize the time-to-market 
  effort to develop new applications : the source code is independent of the language 
  which accelerate considerably the integration on target platforms. 

  Some of the basic features of this product are : 
  - Continuous / Isolated word recognition
  - Push To Talk or Speech Detector
  - small to medium vocabulary size
  - 11 languages available

  - CPU Requirements : from 80 to 160 Mips depending on the sampling frequency 
    and the platform.
  - ROM code : 70 Kbytes
  - RAM data run time : 500-700 kbytes
  - ROM data : 550 Kbytes
  - Available platforms : 
           single processor RISC : ARM, MIPS, SH3, SH4, etc...
           DSP : TMS 320C55x, ... On going development ...
           dual core processors : OMAP (Texas Instrument)
  - Available OS : Pocket PC, Linux, SYMBIAN (under development)

  More Information