Another interesting data type is audio files. Models that convert speech in to text or classify audio sounds take as input audio files. If you want to work with audio files, then you will need the librosa module. There are many ways to treat an audio file; we can convert it into a time series and use a recurrent network. Another way that has given good results is to use them as one-dimensional or two-dimensional patterns, and train a CNN to classify them. Some good papers that adopt this approach are as follows:
-
Hershey, S., Chaudhuri, S., Ellis, D. P., Gemmeke, J. F., Jansen, A., Moore, R. C., and Slaney, M. (2017, March). CNN architectures for large-scale audio classification. In Acoustics, Speech, and Signal ...
