Allowed file types:jpg, jpeg, gif, png, webm, mp4, swf, pdfMax filesize is 16 MB.Max image dimensions are 15000 x 15000. You may upload 5 per post.
File: 658d393ab8c7a37⋯.png (595.16 KB, 738x531, 82:59, spectrogram_no_peaks.png)
File: dc7a73da736575d⋯.png (497.54 KB, 738x440, 369:220, spectrogram_peaks.png)
File: 159000ae6dea84c⋯.png (405.17 KB, 628x558, 314:279, spectrogram_zoomed.png)
Music Matching Thread Gastarbeiter 02/12/19 (Tue) 12:10:30 No.98937
ITT: We discuss ways of searching audio clips
Idea 1: Spectrogram fingerprinting
https://oxygene.sk/2011/01/how-does-chromaprint-work/
(Used for AcoustID)
Idea 2: Peak Value fingerprinting
https://willdrevo.com/fingerprinting-and-audio-recognition-with-python/
https://github.com/adblockradio/stream-audio-fingerprint/blob/master/README.md
https://medium.com/@treycoopermusic/how-shazam-works-d97135fb4582
https://www.ee.columbia.edu/~dpwe/papers/Wang03-shazam.pdf
Gastarbeiter 02/13/19 (Wed) 06:43:30 No.98959
bump for audio sciences
Gastarbeiter 02/13/19 (Wed) 13:08:25 No.98963
The first one is how I imagined they all worked. A bit like image similtude comparing some heavily downscaled and posterized versions.
Gastarbeiter 02/13/19 (Wed) 14:27:48 No.98965
>>98963
Compare Idea 1 and Idea 2, which is better?
Gastarbeiter 02/13/19 (Wed) 18:16:26 No.98968
>>98965
I prefer the first, but I honestly think this is a (possible easy) problem for machine learning.