Project 02: Speech Recognition and Machine Learning | Arduino Nano 33 BLE Sense Roadtest

RE: Project 02: Speech Recognition and Machine Learning | Arduino Nano 33 BLE Sense Roadtest

gpolder — Thu, 01 Apr 2021 13:42:34 GMT

did you try an FFT as preprocessing step?

RE: Project 02: Speech Recognition and Machine Learning | Arduino Nano 33 BLE Sense Roadtest

DAB — Wed, 31 Mar 2021 19:25:20 GMT

There are a lot of variables involved with getting audio recognition correct.

I remember back in the 1980's we had a system that worked well for most of us, but our engineer from Georgia could not use it.

When we tuned it to work for her, it would not work for any of the rest of us.

DAB

RE: Project 02: Speech Recognition and Machine Learning | Arduino Nano 33 BLE Sense Roadtest

ralphjy — Wed, 31 Mar 2021 14:28:00 GMT

When you trained and built the Edge Impulse model, what accuracy did it indicate? Can you share your impulse design and the resulting Confusion matrix. Just interested as I am learning TinyML also. If your model trained okay, you may have a deployment problem. Did you try live classification with your model?