FFT and Music Comparison

Question

I'm trying to play around with some music clustering algorithms, and I thought that using a feature vector consisting of basically a discretized fft (like discretize the frequencies) would be a good similarity measure. Would this even be useful? Do people know what some good audio similarity measures might be?

score 3 · Accepted Answer · answered Mar 05 '12 at 08:44

3

First of all, you need to decide whether you want fingerprinting (i.e. identity except for some distortion) or similarity (but not identity!) measures.

Also have a look at MFCC, bark scales and so on. There is plenty of literature out there. Go to Amazon, and grab a dedicated book on this topic.

answered Mar 05 '12 at 08:44

Has QUIT--Anony-Mousse

76,138
12
138
194

I'm looking for similarity, but yes I've been reading up on the literature now much more. Thanks for the tips! – themaestro Mar 06 '12 at 01:28
I've you find something that lives up to this, keep me updated. The stuff I've tried so far was not too convincing for similarity, but primarily fingerprinting. There is an mpeg-7 feature extractor somewhere available, but even for those features that are meant to be used for similarity, it was unclear to me how to compute an actual similarity. – Has QUIT--Anony-Mousse Mar 06 '12 at 06:59

score 0 · Answer 2 · answered Mar 07 '12 at 23:23

0

You can use a hierarchical cluster like a kd-tree or a hilbert curve before you discretize. A cluster reduces the dimension complexity and change the order of the input while a fft just transform it to waves.

answered Mar 07 '12 at 23:23

Micromega

12,486
7
35
72

FFT and Music Comparison

2 Answers2