Wrong tonic identification result returned in joint analysis

[sertansenturk]

When I call: docserver.file_for_document(mbid, 'jointanalysis', subtype='tonic'), the docserver returns the tonic identification result of audio analysis instead.

This occurs in the following recordings:

43161191-a382-4f4a-9601-828ada3b711b
bee6ac55-19d8-4bdf-891e-ea2c8cc91837
fad224f8-0973-49ad-9c14-7626b1556411
ff8836b8-ee9b-4351-b30c-03b2aac0dc25
8ff4d355-dd46-4074-b266-aeacabaf875a
1d7aa01a-8bf2-43e1-b972-714699db563d
1035e5bd-5f53-4a81-ad66-f861a51d8c4e
820b1b89-3596-42ec-b1b7-26d0974b0dd2
3601efbc-b05b-4f54-ba07-584a842de72e
5acbd63f-63fa-4070-a2a2-79e2940de768
6d1b0211-153e-4b00-ba3f-b48714216701
fad224f8-0973-49ad-9c14-7626b1556411
ffefc26e-4854-4af1-83b9-ac4e773dfc75
36a7a9a1-faac-439d-90bb-117597e13ffc
12567f93-4ae7-4198-a55b-1422b7a3935c

[andrebola]

This Mbid are not recordings they are works, so they don't have audioanalysis nor jointanalysis.

[sertansenturk]

Ohh i pasted wrong mbids. I'll correct them and notify you as soon possible. El 22 dic. 2016 12:52, "Andres Ferraro" <[email protected]> escribió:

…

This Mbid are not recordings they are works, so they don't have audioanalysis nor jointanalysis. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#413 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AEVI4Vvvjlwf5h0kuuAceahQUdyjGcZyks5rKmSNgaJpZM4LTgqC> .

[sertansenturk]

@andrebola, sorry I couldn't write earlier because I was traveling. I couldn't run the checks for the complete audio collection yet, but I wanted to provide an example in the meantime: http://dunya.compmusic.upf.edu/document/by-id/5e2153e4-03fc-47a7-b91c-67c085ada644/jointanalysis?v=0.1&subtype=tonic

This is the output of audioanalysis tonic identificaiton method, which somehow propogated into the jointanalysis: http://dunya.compmusic.upf.edu/document/by-id/5e2153e4-03fc-47a7-b91c-67c085ada644/audioanalysis?v=0.1&subtype=tonic

[andrebola]

@sertansenturk looking at the extractor, I guess they have the same value due to some failure on jointAnalyzer. Can you check if this code it's right?
https://github.com/MTG/pycompmusic/blob/master/compmusic/extractors/makam/jointanalysis.py#L95-L100

or should we store the output of summarized_features?

[sertansenturk]

I found out that the problem is in the score-informed tonic identification in joint analysis. Somehow the tonic value is not returned but the alignment etc. are done without a problem. It's quite strange.

I couldn't pinpoint the problem yet because the problem seems to be on the MATLAB side. I'll probably investigate it thoroughly next week, when I return to Barcelona.

[sertansenturk]

Since it's a tomato related issue, I moved it to: sertansenturk/tomato#73