I’m wondering if there is a way to use the library generated with edge impulse for voice recognition with an STM32L432KC.
Yes, just export as C++ Library or STM32Cube.AI pack and that gives you a library that will run on the target. You’ll need to hook up your microphone but I’d recommend reading through https://docs.edgeimpulse.com/docs/running-your-impulse-locally-1 as a start, once you’ve that working connecting the sensor is not a lot of work.
Thank you for your feedback.
@Saidane as @janjongboom mentioned, you can definitely run Edge Impulse C++ libraries on most other STM32 parts (I think they need to be Cortex-M4 or better/more powerful, though). I love the L432KC–it’s my favorite Nucleo right now. I put together a demo (and guide) on how to run an Edge Impulse library on it here: https://github.com/ShawnHymel/ei-keyword-spotting/tree/master/embedded-demos/stm32cubeide/nucleo-l432-keyword-spotting. Hope it helps!
@ShawnHymel thank you for your help .
@ShawnHymel thank you for the work you have done with STM32L432KC, i really appreciate.
In my case i want to use the Microphone Amplifier with AGC and Low-Noise Microphone Bias MAX9814, I will try to make the necessary changes.
@Saidane Glad it helps! Let me know if you get the AGC microphone working–I have been wanting to do AGC with the project I listed but haven’t had time. My thought was to use a basic I2S microphone and then implement per-channel energy normalization (PCEN) as per this paper (http://www.justinsalamon.com/uploads/4/3/9/4/4394963/lostanlen_pcen_spl2018.pdf).
@ShawnHymel Thank you for your encouragement. I’m also wondering if you know how to implement a library generated with Edge Impulse via Mbed on a STM32L432KC and ADA3421.
Thank you in advance.
@Saidane I don’t know mbed very well, so I won’t be able to help here, sorry! I imagine that you’d be able to import the C++ library and just make calls to any of the available functions in the library. I just don’t know how you link stuff in whatever mbed build system you’re using (online IDE, mbed IDE, PlatformIO, etc.).
@Saldane your best bet would be to find a driver for the
ADA3421 and making sure you can capture audio from it. Afterwards hooking it up to the inferencing library is straightforward.