Dear Francesca: The data used in this post is pretty simple. For example, one sample has 500 data points from a sine function and 1 data label from 0/1/2. The real world dataset, e.g., urban sound (https://urbansounddataset.weebly.com/) or music genre (http://marsyasweb.appspot.com/download/data_sets/), are more complicated than my toy dataset. I don’t know what kind of waveform/signal you have. You might check machine learning/genre recognition under https://musicinformationretrieval.com/index.html for reference.