Agronesia

On a previous post I discussed about the results of a Windows BPM finder program MixMeister BPM Analyzer. It sometimes produces values that are 2 times or half the seemingly actual value. And then the doubt grew on me, “Is that program even accurate at all?”. So, I was set to program my own BPM finder.

But I don’t know any sound programming stuffs, so I went for the easy path of a human-assisted program in which the algorithm is very intuitive. A human would tap the beats while playing a song and the program will do the simple mathematical calculation (it’s “beats per minute” for a thing, so it’s basically just a division).

(more…)

MixMeister BPM Analyzer - analyzing DOES take time

The BPM (Beats per Minute) of a song can be stored as an ID3 tag on your MP3 files. In essence, the BPM is the tempo of a song which indicates how fast or slow a song is.

Programs like iTunes can display the stored BPM number and utilize it. If your song collection has BPM value, you can for example make a playlist for the fast songs and another for the slow songs. The fast songs can then be copied to your MP3 player for your gym activity, for example.

The problem now is that this value is initially empty. We can even insert a random number on programs that supports MP3 metadata editing. Luckily, there are programs that analyze BPM and store the computed value on the files. One such program is MixMeister BPM Analyzer which is free to download. Using this freeware is very easy, you just need to drag and drop the files or folders you want to analyze. It’s computationally quite heavy so if you have tons of files be prepared for a long wait.

(I wonder whether there is an open-source program for this purpose)

The result in iTunes after the BPM is analyzed

I tried searching for such programs because I wanted to know the BPM of songs I want to try playing on the keyboard. The calculation output is quite interesting. For most files, the program seems to be accurate. However, for some files the software computes a BPM which is actually 2x faster than the seemingly real value or else half of it. For example, the program gives 82.42 BPM for a song (Berryz Koubou’s “Gag 100kaibun Aishite Kudasai“) despite it being a fast one, so the double value 163.84 BPM seems to suit more. There’s also a slow song (Morning Musume Sakura Gumi’s “Sakura Mankai“) which is given 160.03 BPM, but clearly 80.015 BPM is closer to the real answer.

But theoretically, a musical score written in x BPM can be easily rewritten in 2x BPM or 0.5x BPM. We just need to double or halve every note’s duration. Of course, by this modification a note might be moved to another measure. In a sense the program is correct.

What kind of baffles me is that I can probably sense the beat of a music intuitively, but I wonder how one makes an algorithm that can find BPM. Ah, sound processing, an area where I have practically null experience.