I've audio files say "left.wav", "right.wav" and so forth, I want to create a model which takes audio as input and output label "left" or "right" etc.
Question
How do I feed my raw audio to my neural network ?
Asked
Active
Viewed 197 times
1 Answers
0
The scipy.io.wavfile.read() function will return the sample rate and the whole audio in a numpy array.
You can then feed that to your network.
import scipy
rate, numpy_audio = scipy.io.wavfile.read( "left.wav" )
If you want to do speech recognition, check out DeepSpeech, it's a large project, but you can probably get some good ideas there.
For a simpler intro, Tensorflow has a Simple Audio Recognition tutorial.
To generate audio, you might want to consider WaveNet - this is one particular implementation.
Peter Szoldan
- 4,792
- 1
- 14
- 24
-
That's it?Nothing more? – martian1231 Apr 17 '18 at 00:33
-
Is it too easy? :) – Peter Szoldan Apr 17 '18 at 00:33
-
Are you trying to do speech recognition? – Peter Szoldan Apr 17 '18 at 00:36
-
Added some links for you in my answer – Peter Szoldan Apr 17 '18 at 00:51
-
hi, Thankyou so much. Yes, I'm just trying to create a basic speech recognition model using tensorflow. :) – martian1231 Apr 17 '18 at 01:11