HVO_Sequence

[behzadhaki]

Feature Implementation 1

• [x] pickling should only save/load only hit events (tuples of (step_idx, voice_idx, vel, offset) for all hits))

Feature Implementation 2

• [x] implement shift by num steps
• [x] implement concatenate to append to two HVO_Instances
  • [x] zero append to ensure length of first one is multiple of n_beats
  • [x] add an empty version of the first sample to the beginning of the second one (update qpm and meter timings for second one as well)

[behzadhaki]

• [x] Import/Export/Synthesis for multi meter, multi beat_div, multi tempo audio should be tested

[behzadhaki]

• [x] segmentor implementation
  • [x] Segment HVO into consistent segments
  • [x] Segment HVO into equal length Segments (length can be either n_steps/n_beats/n_bars)
• [ ] Downsample a long performance into N segments
  • [ ] use cross-segment cosine distance to rank diversity of loops --> sort accordingly and return N equally distanced samples
  • [ ] use weighted note density (total notes / Num voices) instead of similarity

[behzadhaki]

• [ ] Heatmaps: Think how to create heatmaps using all voices combined (maybe scale each voice heatmaps by a factor denoting the proportion of hits in a specific voice divided by hits in all voices)