How to change test signal in "Denoise Speech Using Deep Learning Networks"?

I am able to successfully run the example "Denoise Speech Using Deep Learning Networks" without any modifications. In the section 'Test the Denoising Networks' the code currently chooses a sample from the test dataset within the Mozilla Voice dataset. I would just like to be able to choose a specific wav file that I have located on my computer as the test signal. How do I do this please?

Link to the example here: https://uk.mathworks.com/help/signal/ug/denoise-speech-using-deep-learning-networks.html

Thanks!

Answer 1

[jibrahim]

Hi Charlotte,

You can try changing the first line of code in that section to something likeL

adsTest = audioDatastore(myfolder);

where myfolder is the folder containing your clean test signal. The rest of the section should work.

This network was trained on washine machine noise only, so I do not expect it to perform well on other types of noise. The example section uses "WashingMachine-16-8-mono-200secs.mp3" as a source of noise, but if you have another noise file, just replace the name with yours.

[Charlotte Jones]

Thanks! This worked for me.

[Tom Cipollone]

I have downloaded the zip file and am running the denoise example except for the fact that the file "WashingMachine-16-8-mono-200secs.mp3" is missing from the archive.

Where can I find this file?

Thank You

Tom

[jibrahim]

Hi Tom,

The file WashingMachine-16-8-mono-200secs.mp3 ships with Audio Toolbox, so it should be on your path if you have the product installed.

[amina]

svp comment trouver d'autre fichier d'audio ou de bruit MP3 hors celui utiliser dans cette exemple . je vous remercie d'avance