Hello, im trying to replicate this project and having some trouble on my model. The learning block i should use is Transfer Learning (Keyword Spotting) but it requires 16k hz sample rate and i can only use 4k. I’ve tried Classification learning block but after uploading the data to the pico the values are random. The electronic configuration is like in photos. Any help would be appreciated!
Hello @Felix61287,
Indeed, the transfer learning keyword spotting model has been pre-trained with 16kHz audio.
I’d recommend changing your model to a custom classifier.
See Responding to your voice | Edge Impulse Documentation.
Both MFE and MFCC should work for the pre-processing. Feel free to try the EON tuner to test different impulses.
Best,
Louis