Speech Recognition Performance RRS feed

  • Question

  • Hi - I noticed that the speech recognition performance is between 3.5 and 8 milliseconds to achieve a recognition result, but it performance seems uneven and it takes too long to achieve a recognized result for my needs.  Is there a way to be able to tune how quickly it takes the engine to achieve a recognition result - even at the expense of the quality of recognized speech?  Currently using the v1.x sdk, but open to upgrading to v2.x if required.



    Tuesday, April 14, 2015 2:48 AM

All replies

  • Speech platform is a separate technology from Kinect SDK. You will be using the same recognition libraries with Kinect v1 or v2 except for the language packs.

    I'm afraid I don't know any way to increase recognition speed. I think there is no other specific forum for Speech platform neither.

    Tuesday, April 14, 2015 8:29 AM