尝试使用法国声学模型

Question

I'm trying to use a french acoustic model with OpenEars. The answer of Ramshad is very interesting. But I have downloaded French Acoustic Models and I can't figure out how to use it...

I have made a bundle with those files and I have added it to my project and I call generateLanguageModelFromArray method. But when I open the generated .dic file, the phonetic chars seem to be wrong. If I try to speak, it doesn't recognize anything.

The log is the following : "Pocketsphinx has detected a period of silence, concluding an utterance."

Could someone explain me a bit better how should I use it ?

Thanks

Answer 1

OpenEars developer here. Since OpenEars doesn't support French, you can't use LanguageModelGenerator to generate language models or grammars, since it only supports the phonemes in English and Spanish.

You can create a language model and phonetic dictionary that work with your acoustic model by hand by referencing the phonemes which are visible in the phonetic dictionary that ships with the French acoustic model from the CMU site, but you should test those three elements out in a Linux (or Linux VM) Pocketsphinx install and verify that they work perfectly in Pocketsphinx before putting yourself through the stress of troubleshooting them in a framework which doesn't support the language and assumes that it has a language as input that it knows about.

At the point that it's working well outside of OpenEars, please feel free to come over to the OpenEars support forums and get assistance from me for bringing your known-working acoustic model, language model and phonetic dictionary into OpenEars. It would also be helpful to take a look at how OpenEars logging works, so you can provide informative logs there. In the case you have right now, an informative and probably pretty specific error about the acoustic model and language model is probably being logged in the verbose logging output for you to see in your troubleshooting.