Issue running voice model with microphone on arduino-nano33-ble


I have a project where I have tried three different words to create a model and upload on the Arduino nano 33 BLE. It was working fine initially but recently I have an issue.

When I export the arduino .zip library and run it in contnuous mode the classification gets fixed to specific value and it does not change evenif there is a noice or specified word in the model.

I suspect that the Microphone is not working hence I uploaded the data-collection firmware and recored the voice. The microphone can record the voice and I can hear it.

I suspect the problem is in the library but the same library worked for sometime in the past!!

Hi @ujjwalrathod00700,

Have you modified the continuous mode sketch?
Also using the default firmware, are you able to classify correctly in the Live Classification?


Hi @aurel

Yes, I had modified the code once, but then I unlinstalled the library and reinstalled the library. But also with the original code the classification is not correct.

In live classification also there is not correct classification, but previously it was.

Hello @ujjwalrathod00700,

I am not entirely sure but could you try changing the PDM gain from the microphone in your arduino code to see if you have better results?

Please see: Arduino nano 33 BLE inferences make no sense