Im using a gan which generates music. I have converted the wav files into Mel Frequency Cepstral Coefficients by using python_speech_features’s.mfcc. the code for that:
signal, rate = librosa.load('drive/MyDrive/Proj_NN/samples/'+wav_file, sr=8000)
signals_1[k] = signal
# nfft = (8000/second) / (40 intervals/sec)
# 40 intervals/sec = 25 milliseconds (resolution)
# nfft = 200 samples/interval (size of the window)
## note that nfilt was 26 so we throw away half
mfc= mfcc(signal[:rate],rate, numcep = 13, nfilt=26, nfft=200).T
mfccs_1[k] = mfc
this works fine for me it gives me the mfcc file and the correct mfcc plot.
How do I convert the mfcc file back into a wav so I could listen to it?
I need to know how the conversion from mfcc to wav occurs as the output of my gan is an mfcc file/ image so i would have to listen to the audio to evaluate my model.