Is sincnet possible to convert back to original wave?

mravanelli / SincNet

SincNet is a neural architecture for efficiently processing raw audio samples.

MIT License

1.14k stars 263 forks source link

Is sincnet possible to convert back to original wave? #16

Closed r06944010 closed 5 years ago

r06944010 commented 5 years ago

After using first CNN layer (sincnet) to process the waveform , is it possible to convert the feature extracted back to the original waveform ? Thanks

mravanelli commented 5 years ago

The current version is designed for classification rather than for generation purposes. I'm actually thinking some possible ways to use the "SincNet" idea to generate an audio waveform with fewer parameters, but so far I don't have results on that...

mravanelli commented 5 years ago

In you case, you can try to add an autoencoder that should be able to recover an approximated waveform from the hidden representations...

r06944010 commented 5 years ago

Thank you very much!