Song peak detection (bpm counter)

[nuno12345]

Hello, I'm trying to develop a BPM counter for music beatmatching but I can't find references anywhere so I started a new thread.

My guess is that I've to make a lowfreq filter first for "beat" detection, then detect each peak and the time between those peaks.

Any help?

Thanks

[bkamen]

I did something similar recently. (VU meter. Fun!)


Do you have your filter/peak detector circuit yet?

-Ben
_________________
Dazed and confused? I don't think so. Just "plain lost" will do. :D

[nuno12345]

well i think for filter we can simply use an RC circuit, low pass filter quite simple....dont have a peak detector but probably i can use ADC right?

[bkamen]

I would make sure to use a good low pass with an op-amp.

You'll need the drive of an op-amp to make most ADC's happy.

If you look on Microchip's website, they have a free download for a "filter designer" called FilterLab.

In a sense, you could just as easily use a peak detector on a comparator as you could an ADC to figure out at what point you have what level.

But either way, you still need to start with a good filter that takes out the stuff that can interfere with the signal you want to measure.

-Ben
_________________
Dazed and confused? I don't think so. Just "plain lost" will do. :D

[vascotech88]

Anyone found solution already?

[bkamen]

I did a peak detector for another project (the VU meter) that worked quite nice.

The software just drives a VU display though.. it's not trying to detect peaks.

So I don't know if that's a solution for you.

I used a low-pass filter with an active rectifier.

-Ben
_________________
Dazed and confused? I don't think so. Just "plain lost" will do. :D

[vascotech88]

Hello, again.

After this post here today, i started working in MATLAB to test some filter and peak detections and later write in C and put in a PIC.

My idea is this:
-We have a sound from 0~20KHz
-Since we need to detect beats, we need only arround 500Hz of max frequency.
-In my case, i will use 2KHz of ADC conversion frequency in the PIC, which will make 1KHz of max frequency, and its enough
-Then i will make a simple convultion filter with 16 points, wich is enough to not start having echo after the calculations
-Then we can apply to all points: x^4
-Then we just need a threshold value
-After... well i will do tomorrow, but we need to filter the beats, give some time to wait for the next one, and measures, automatic gain and so on...

(NOTE: Of course we need normalizations and so on... i added them in the matlab file)
(NOTE2: I am just planing to use a voltage divider, condensator and 2 resistors to put mean value of sound in 0.25V)

Picture here:

[vascotech88]

I forgot to tell, the song is Into The Future (The Aston Shuffle Remix) :P

[vascotech88]

Here I am back again.

I made some tests in Proteus about the input circuit... At a first approach I didn't want to use opamps and so on, yes, you may ask about the input power, well this circuit is to work with high sound power, not a regular headphones output. That's why the idea of the voltage divider. Anyway, after some mins playing with Proteus I got this:

[vascotech88]

Finally i spent some more time improving the MATLAB C file.
Its not finished yet:
> I need to make the auto gain code, which is not easy to make in MATLAB because i am not running in real-time, but i will try.

Here is the latest picture:



I will upload the files soon
_________________
-----
Vasco Jose Dias Baptista
http://vascotech88.pt.vu/

[vascotech88]

Check my latest update!

http://www.youtube.com/watch?v=ALCoSOcjPY8
_________________
-----
Vasco Jose Dias Baptista
http://vascotech88.pt.vu/

[vascotech88]

CHeck my website, i made the first working prototype (more or less) :p
_________________
-----
Vasco Jose Dias Baptista
http://vascotech88.pt.vu/