Disclaimer : Information provided here may not represent the standpoints of Carnigie Mellon University or CMU Sphinx Group.

Sphinx Open Source Projects for Researchers

Completed Projects

If you are interested in any of the following projects, please send mail to me at archan at cs dot cmu dot edu

Sphinx Model Editor

Project Category: Trainer

Goal:

Although Sphinx uses a very simple data structure in the application and it is very easy for developers to change and use them. There is no easy way for users to change and display the models easily.

Our current goal is to expand printp and cp_parm.

Sphinx 4's trainer.

The success of Sphinx IV shows that Java is a competent language for high speed and high performance computation. Yet usage of Sphinx IV is still limited by the fact that there exists no Java version of trainer. We hope that someone can help us to port SphinxTrain to Java and allow Java user to enjoy the use of high performance acoustic modeling tools.

Completed Projects

SphinxTrain Command-line Cleanup

Project Category: Trainer

Goal:

(New at 20051203) Sorry, we haven't updated this page for a while. The whole command-line clean up process is completed. The new SphinxTrain code could be download from Sourceforge CVS.

SphinxTrain is a powerful acoustic modeling toolkit developed by researchers in CMU. However, its current development is hindered by difference of the commandline arguments of different tools. In this project, we are seeking for ways to clean up the difference of the codes.

We also want to include feature such as -help and -example to each toolkit.

Tool which is known to be not conform to the standard commandline outputs are

QUICKCOUNT

wave2feat

Tools which are known to be more suitable merge together

mk_s2cb, mk_s2hmm, mk_s2phone, mk_s2phonemap, mk_s2sendump, mk_s2seno, mk_s3gau, mk_s3mixw, mks_s3tmat are better to merge to be a tool called model converter