Write a function to process the audio object $($torch tensor with its sample rate$)$ to the

augmented mel spectrogram $($torch tensor$).(15\%)$

Input parameters contain an audio object, the re$-$sample rate, the density of both

masking, the number of frequency masks and the number of time masks

Use the re$-$sample rate new$\_$sr to adjust the sample rate of the audio object

No need to change the audio object to mono or stereo channels

Construct the mel spectrogram $($torch tensor$)$ based on the adjusted audio object

Use max$\_$mask$\_$pct$,$ n$\_$freq$\_$masks and n$\_$time$\_$masks to create the frequency

masks and the time masks on the mel spectrogram

Return the augmented mel spectrogram $($torch tensor$)$

As randomness is involved in masking, your result may be different from the

specification. However, the shape of the augmented mel spectrogram should be the

same.

This problem covers audio signal processing.