I am working with deep learning algorithms like CNN and RNN.I always wonder what is the best way to input wave form type data in to the deep learning algo. I know there are methods like wavelet or mel coefficient. But those things are man made things so like SIFT features in image processing.I'v seen some people take stft images and use them as inputs. What are the other ways?
$\begingroup$
$\endgroup$
Michael R. Chernick
42.9k28 gold badges83 silver badges157 bronze badges
asked Apr 22, 2017 at 13:22
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
You can also put audio/speech into a CNN in raw/time-domain form, as WaveNet has done: https://deepmind.com/blog/wavenet-generative-model-raw-audio/
