none
Kinect Echo cancellation and Noise Suppresed output RRS feed

  • Question

  • Hi,

    I have a question on whether I can record the output after the 4 mic captured data (using the kinect sensor) has gone through noise suppression and echo cancellation (before it is sent to the speech recognition block) -From the limited time I have spent browsing the website, I have found that You can record all the 4 mic signals (Raw capture) but I have not been able to find out how to record the noise suppressed/echo suppressed output

    I am interested in recording the noise suppressed and echo suppressed output -is this possible ? 

    My goal is to benchmark the Kinect's echo and ambient noise suppression against other echo and ambient noise suppression algorithms available.

    Monday, November 7, 2011 3:13 PM

Answers

  • Sorry about that, Labaqu. The DMO output is single-channel, not 4-channel, so there is currently no way to use DMO for echo cancellation and/or noise suppression functionality and also get 4 channel audio.

    Eddy


    I'm here to help
    Wednesday, December 14, 2011 11:21 PM

All replies

  • You should look at the MicArrayEchoCancellation sample application, which does exactly that.

    Eddy


    I'm here to help
    Monday, November 7, 2011 9:29 PM
  • Hi Eddy,

    Are you saying that in MicArrayEchoCancellation application, as long as we change the number of channels to four and set the bufferCount parameter in pDMO->processOutput to four as well, we can get recording from four channels?

    Thanks!

    Tuesday, November 8, 2011 9:07 PM
  • If I remember correctly from a previous post, the MicArrayEchoCancellation sample outputs a WAV file that contains four audio channels, one for each microphone. Search around this board, you'll find the post I'm thinking of...

     

    EDIT: Here it is... http://social.msdn.microsoft.com/Forums/en-US/kinectsdkaudioapi/thread/40388b70-a845-4b9b-9bed-1cebed04653d

    • Edited by ChrisCicc Tuesday, November 8, 2011 11:20 PM
    Tuesday, November 8, 2011 11:19 PM
  • Hi ChrisCicc,

    But that is for AudioCaptureRaw which uses WASAPI to get the four channels. In MicArrayEchoCancellation sample code, it uses DMO object. I am not sure whether this object is capable of extracting four channels at the same time.

    Wednesday, November 9, 2011 2:25 PM
  • Sorry about that, Labaqu. The DMO output is single-channel, not 4-channel, so there is currently no way to use DMO for echo cancellation and/or noise suppression functionality and also get 4 channel audio.

    Eddy


    I'm here to help
    Wednesday, December 14, 2011 11:21 PM