Unexpected drop in loudness_decr ramp
anusfoil opened this issue · comments
When computing the dynamics direction feature 'loudness_direction_feature.loudness_decr' for Chopin Op.10 No.3, there is an unexpected drop of the ramp value from 4 to 0 but there is no clear indication of the change on the score. To reproduce the problem:
score = pt.load_musicxml('Chopin_op10_no3.musicxml')
na = pt.musicanalysis.note_features.compute_note_array(score, feature_functions=['loudness_direction_feature'])
na['loudness_direction_feature.loudness_decr']
Output is as follows (not sorted by time). From n155 to n156, there is a sudden drop from value 4 to 0, which corresponds to no loudness decrease change at this point in score. Expected behavior: no change in the ramp value and 4 is kept towards the end of the piece.
array([0. , 0. , 4. , 4. , 4. ,
4. , 4. , 4. , 4. , 4. ,
4. , 4. , 4. , 0. , 4. ,
4. , 4. , 4. , 4. , 4. ,
4. , 4. , 4. , 4. , 4. ,
0. , 4. , 4. , 4. , 4. ,
4. , 4. , 4. , 4. , 4. ,
4. , 4. , 0. , 4. , 4. ,
4. , 4. , 4. , 4. , 4. ,
4. , 4. , 4. , 4. , 0. ,
4. , 4. , 4. , 4. , 4. ,
4. , 4. , 4. , 4. , 4. ,
4. , 0. , 4. , 4. , 4. ,
4. , 4. , 4. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0.50000006, 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0.50000006,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0.50000006, 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 1. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 1. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 1. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
1. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 1. , 1. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 1. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 1. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
1. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 1. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 1. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
1. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 1. , 1. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 1. ,
0. , 0. , 0. , 0. , 0. ,
0. , 0. , 0. , 0. , 0. ,
1. , 0. , 0. , 0. , 0. ,
0. , 1. , 1. , 1.5000001 , 1.5000001 ,
0. , 0. , 1.5000001 , 2. , 2. ,
2. , 2. , 2.25 , 2.25 , 2.5 ,
2.75 , 2.75 , 2.75 , 0. , 3. ,
3. , 3. , 3. , 3. , 3. ,
3. , 3. , 3. , 3. , 3. ,
0. , 3. , 3. , 3. , 3. ,
3. , 3. , 3. , 3. , 3. ,
3. , 3. , 0. , 3. , 3. ,
3. , 3. , 3. , 3. , 3. ,
3. , 3. , 3. , 3. , 0. ,
3. , 3. , 3. , 3.5 , 3.5 ,
3.5 , 3.5 , 4. , 4. , 4. ,
4. ], dtype=float32)
Score is attached with red marking of beat n156. Conjecture is that the end of a crescendo (loudness_incr) direction affects the ramp value of loudness_decr.
Dynamic Features do possibly not work as expected. I fixed the ordering of the features by adding a sorting step at the end of compute_note_array, but the increment and decrement features still stay constant until the next constantloudnessfeature (the mf in the score) and then drop to zero. Not sure if this is intended behavior, maybe they should drop right at the end of the marking instead, the stacked features seem a bit weird. I'll tentatively link a PR.