so we'd need to implement #115 first

see
https://github.com/mvansegbroeck/mupet/blob/ae7dc9e68ec74bfe7d2f5c5214ae32489a7088e9/core/audio/compute_musv_segment.m#L57
that uses energy E as returned by this function
https://github.com/mvansegbroeck/mupet/blob/ae7dc9e68ec74bfe7d2f5c5214ae32489a7088e9/core/audio/compute_musv_segment.m#L162

This would let us replicate part of AVA paper where they extract all syllables from MUPET dataset, see figure 2
https://elifesciences.org/articles/67855/figures#content