Closed janden closed 6 years ago
I agree, though in back and white this might be tricky. Let 's assign that to gregoire ;)
Instead of colors, we can have a "track" below that indicates the current cluster (a-d) and have a legend somewhere.
good point
that being said, we need room for that in the paper if gregoire is willing to do it
This is maybe beyond what we have time to do, but I'm putting this here just in case we do.
What about taking an auditory scene (represent as a scalogram), and label it according to the clusters found in the k-means step. Let's say we pick a small number of clusters (4). Can we identify the background vs. sound events that we're interested? Could be a fun motivating figure to put in the clustering section.