[Kde-accessibility] Simon: Noise and Echo Cancellation Tests

[Peter Grasch]

Hi Jon,

I set up a testing environment on the server for you.

Please find the steps below to use it to test the performance of one
preprocessing strategy.

1. Go to /speech/en/am/sphinx/voxforge_ted_vctk_small_nc_test
2. Remove the folder "samples" and it's contents
3. Copy the folder "samples.stock" to "samples"
4. Run the currently tested cleaning algorithm on all .wav files in the
new "samples" folder.
5. Go to /speech/en/am/sphinx/voxforge_ted_vctk_small_nc_test/test
6. Remove the folder "samples" and it's contents (same as before, only
in the "test" sub folder)
7. Copy the folder "samples.stock" to "samples"
8. Apply the same algorithm as in (4) on the test samples
9. Go back to /speech/en/am/sphinx/voxforge_ted_vctk_small_nc_test
10. Run "sphinxtrain run". This will take a while. Check the debug
output for errors.
11. Go back into the "test" sub folder
12. Run "./test.sh"
13. Note down the details of your new model and the results you achieved
with it on the spreadsheet:
https://docs.google.com/spreadsheet/ccc?key=0AuD4Nshr5xg5dHV0ODdtb0VnOUVrMkl3bzVOcmhwdUE&usp=sharing

If you have any questions, I'd be happy to answer them.

Best regards,
Peter