1

I'm trying to develop an iOS app using speech-recognition for french language, but have been unsuccessful until now. I tried using the openEars framework, which worked great for english language, but doesn't support french. I used some info from this link.

If anyone know a solution it would be awesome.

Thanks

Prasad G
  • 6,702
  • 7
  • 42
  • 65
Leguman
  • 2,104
  • 1
  • 18
  • 22
  • solution to what exactly..? getting the framework to support french? – Just a coder Mar 14 '13 at 12:03
  • Speech recognition actually must be supported on iOS through openEars. The text-to-speech does not. The models and everything else is availble on CMUSphinx website, you can just use them. – Nikolay Shmyrev Mar 14 '13 at 13:32
  • @NikolayShmyrev I tried to use the CMUSphinkx langage model direclty [link](http://sourceforge.net/projects/cmusphinx/files/Acoustic%20and%20Language%20Models/French%20Language%20Model/) but I think it is too big (62K words) so it doesn't recognize anything – Leguman Mar 14 '13 at 14:26
  • Right, you can prune language model and reduce it's vocabulary to make it smaller. English model is small too. – Nikolay Shmyrev Mar 14 '13 at 14:40

3 Answers3

2

Openers is using English Acoustic and Language Models by default. So it works well with english, but doesn't support french.

You can download French Acoustic and Language Models from CMU Sphinx website

Some good French Acoustic and Language Models are available here

Download & Change your Openears Acoustic and Language Models with this one.

To do this, just replace Openears hub4wsj_sc_8k folder with the downloaded one and replace the .dic and .languagemodel files with the downloaded ones.

If you think , the downloaded French Acoustic and Language Models are too big, then create our own with few words using this documents

Shamsudheen TK
  • 30,739
  • 9
  • 69
  • 102
  • Thanks I didn't know that you had to replace the **hub4wsj_sc_8k** folder with an other acoustic model. It works a little better, finding some words, but it's still too slow. I would have to create my own dictionary. – Leguman Mar 15 '13 at 09:42
  • okay.. creating your own will take too much time. For this better to use Linux system. between 3-4 French Acoustic and Language Models available in CMU Sphinx website. Download and try all of them. Also search Acoustic and Language Models in google. You may get from other websites. Dont forget to replace your .language model file and .dic files along with the hub4wsj_sc_8k folder.or just make sure you are using French .language model and .dic files. finally Clean your project and run. – Shamsudheen TK Mar 15 '13 at 21:51
0

I ended up using this library iPhone-Speech-To-Text for my project.

It use the google voice api, sending a FLAC file of the speech and retrieving the appropriate text.

To make it work for french language, in SpeechToTextModule.m I juste changed the URL

NSURL *url = [NSURL URLWithString:@"https://www.google.com/speech-api/v1/recognize?xjerr=1&client=chromium&lang=en-US"];

to

NSURL *url = [NSURL URLWithString:@"https://www.google.com/speech-api/v1/recognize?xjerr=1&client=chromium&lang=fr"];

Leguman
  • 2,104
  • 1
  • 18
  • 22
-1

The alternative I found is Dragon Speech SDK for Mobile. You simply need to Register With Dragon Mobile and download SDK and Samples.

It Supports English, French, German Speech Recognition. Also it is working very accurately than openEars. I have tested this for English and its working very well.

Here is screen for that sample which supports Recognition in mentioned language

enter image description here

Note

The SDK is free to download and use for 90 days, but...

βhargavḯ
  • 9,786
  • 1
  • 37
  • 59