Question

[JTrotta]

Very interesting work.
Just a question:
The keras H5 model you create from notebook (i.e. key_word_detection_cnn_success_1.ipynb), uses dataset in csv format.(/data_restaurant/data). Did you create csv files using the same flow in the DSP pipeline?
I mean starting from raw audio files (wav files) and then modified applying the same functions, in the same order, present in the void dsp(float32_t *s1, mode mode) function?

Thank you