The company I work for may soon engage in a project where some level of speech recognition or acoustic analysis might be necessary. We are seeking for advice from someone who has practical experience in that field to:
- Point us out the *open source* software/library that would best suit our requirements.
- Provide us (through a Skype session) an overview of how it works and how the development process would look like.
- The application aims to teach users how to pronounce Te Reo Maori words. It will play a sample utterance of a single word and record the user pronouncing it back. It must compare the user's utterance against the sample pronunciation and provide a score indicating how close/good that was.
Assumptions and constraints:
- We've only got one sample pronunciation of each word (male and female versions) available at the moment. We don't envisage investing more time recording more pronunciations.
- Only open source software and libraries are acceptable.
- The application will only deal with isolated utterances and must be speaker independent. We don't envisage having a training process in the application.
- Comparison could be done either through voice recognition algorithms or acoustic analysis (i.e. formants, frequency, pitch, etc.), as long as the score provided is good/realistic.
We appreciate expertise in the area and are able to contribute financially, to acknowledge the time involved.
If you think you can help us and is interested in doing it, please get in touch with me at teolupus.ext [at] gmail.com.
If not, but you know someone who might be the right person to support us, could you please send this message through?
--- (Edited on 10/4/2011 4:10 am [GMT-0500] by Visitor) ---