Hi
You can see also winpich more useful and Real time speech analysis for research or language teaching
here is the link
http://www.winpitch.com/
The problem with sphinx is that in sphinx4 only the decoding is available, we had to train models in sphinx 3 (for arabic models) and to import them in sphinx4.
I think if I were starting from new, my choice would be Kaldi. Although hardly easy to use, this is the only toolkit I know of that incorporates the very latest advances in recognition technology, such as DNNs, and hence enables you to build a real state-of-the-art system.
http://homepages.inf.ed.ac.uk/aghoshal/pubs/asru11-kaldi.pdf
After implementation of ASR for Arabic language using Sphinx, HTK and Matlab toolbox. We find that HTK provides the best word recognition rate. Notice that HMMs were for phonemes.
You can find in this paper (in french) additional information.
Etude comparative entre les librairies de reconnaissance vocale (NCSP'2014)
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Programming Languages