Make Tf reader with STFT coefficient

[nperraud]

1 quarter of the dataset, STFT with a redundancy of 4

[andimarafioti]

tf.decode_raw doesn't support complex data (as in 3+2j) so that should be handled in some way: (https://www.tensorflow.org/versions/r1.4/api_docs/python/tf/decode_raw). Easiest way is to take the real and the imaginary part as two separate float32 vectors and make tfReader reunite them into complex64 values. Do you think that's a good idea?

[nperraud]

Well it will be simply two channels, one for real and one for imaginary, i.e no complex number.

[nperraud]

Sorry this is the first thing to do.

[andimarafioti]

Like this? https://github.com/andimarafioti/audioContextEncoder/blob/master/utils/tfReaderForStftCoeffs.py

[nperraud]

Probably. Did you also make the file to produce the tfReader file?

[nperraud]

Is this done?

[andimarafioti]

No, I'll address this in the fourth issue.