Music Source Separation with deep learning

JavaScript isn't enabled in your browser, so this file can't be opened. Enable and reload.

IMPORTANT: ONLY ONE OPTION IS CORRECT.

Email *

What's the main contribution of deep learning to the field of music source separation? *

The use deep non-linear models.

The use of the phase as input.

The use backpropagation, it's key to recover the phase from the mixture.

The use the forward pass, that allows the vocals to sound louder.

Which input signal tends to perform better for music source separation? *

Spectrograms

Waveforms

The skip connections in U-net architectures are useful, because.. *

..they allow the gradient to flow, and facilitates learning with stochastic gradient descent.

..they allow to recover at the output the high-frequency details that were present in the input.

..they concatenate the encoder feature-maps to the decoder.

ALL THE PREVIOUS ONES ARE CORRECT!

NMF (Non-negative Matrix Factorization) can be used to perform waveform-based music source separation. *

True, why not?

False, because waveforms are positive and negative (range from -1 to 1).

Which of the following sentences is INCORRECT? *

Masks are used to filter the mixture spectrograms.

Masks are applied directly to waveforms.

Masks can be predicted by a deep learning model.

Submit

Clear form

This content is neither created nor endorsed by Google. Report Abuse - Terms of Service - Privacy Policy