I am one of the maintainers of the Janus Speech Recognition Toolkit. We are currently working on:

1. ‣Parallel decoding in „Ibis“ on multi-core architectures using OpenMP
   

2. ‣Bottle-neck features for front-end processing, particularly for the distant microphone case
   

3. ‣Discriminative training of acoustic models and multi-stream combination
   

4. ‣Acoustic models for dialects, and Arabic
   

5. ‣...