4

I'm developing an Android voice recognition app using RecognizerIntent. The client is concerned with data being transmitted between device and Google's speech recognition server and whether it can be intercepted.

I found very little documentation on what format and how the data is being sent back and forth.

Does anyone know where I can find this information? Is communication between the device and Google speech recognition encrypted/secure in any way?

F L
  • 478
  • 1
  • 7
  • 18
  • 2
    UPDATE: Did more research and found out that the audio is compressed via Speex codec, split down to a smaller byte stream packets that are sent to Google. The results are sent in an array as a JSON object. Question remains, can this be done over secure connection that cannot be intercepted? – F L Jul 17 '12 at 17:44

2 Answers2

2

If your client is concerned about security it's worth to use the speech recognizer which works only on the device and doesn't send data outside. One such tool is Pocketsphinx

http://cmusphinx.sourceforge.net/2011/05/building-pocketsphinx-on-android/

Nikolay Shmyrev
  • 24,897
  • 5
  • 43
  • 87
  • We've looked into that. It does not offer the required accuracy. – F L Jul 17 '12 at 17:27
  • 1
    I'm trying to find more information on Google's speech recognition server functionality. I found this: https://docs.google.com/document/pub?id=1T-UnE6vfrTbtsqPK7mybOswL5Pr_KH2eAi-69jlFHLI , but it mostly talks about implementation, not the inner workings. – F L Jul 17 '12 at 17:33
  • Accuracy may be improved. If you are looking for an exceptional security web service doesn't fit your need anyway though there are secure web services (not google ones) which allow you to send an encrypted data and get confident result. You can also consider an option of intalling secure ASR service at your location. – Nikolay Shmyrev Jul 19 '12 at 10:02
1

See Google's voice search speech recognition service which may be helpful. And, of course, the now famous http://mikepultz.com/2011/03/accessing-google-speech-api-chrome-11/ which shows the workings of the Google Speech recognition in the Chrome Browser.

Community
  • 1
  • 1
Michael Levy
  • 13,097
  • 15
  • 66
  • 100