Does anyone how I can implement Beat-Counting (let my program 'know' wether or not there is a beat at a certain moment and determin the amount of this per minute) if I have acces to fft and pcm data (real-time).
Who is Participating?

Improve company productivity with a Business Account.Sign Up

jhanceConnect With a Mentor Commented:
I've not done this in software but have done it in hardware.  I once built a light system for a dance floor and the requirement was to index a series of lights once for every beat of the music.  I built a low pass filter that rolled off sharply above 100Hz.  This then went to an integrator that needed a signal for about 0.125 seconds before triggering it's output.  This output went to a flip-flop circuit that performed a divide-by-2 function (it looked better to do every other beat that every one).  

Since you are doing this in software and you have the fft, it should be a easy task to run this filter and make it work.  You can even add adaptive features so that as the music changes it can "search" different frequencies and windows for a beat if the default one is lost.
I'm very interested in doing the same thing in my app SuperSonic. If you come up with a successful beat counting routine, would you be willing to put it in a dll and license it to me?


OdieAuthor Commented:

No problem
The 14th Annual Expert Award Winners

The results are in! Meet the top members of our 2017 Expert Awards. Congratulations to all who qualified!

OdieAuthor Commented:

Could you explain that a little more detailed?

I know I have to search for "pulses" in the lower fft-values, but could you be a little more specific in your explanation?  How do I do this?
You pick you range of frequencies and then watch for the presence of a signal for a given time period.  For example, let's say you have broken down your input signal using a FFT into 3 bands, 20-200Hz, 200-2000Hz, and 2000-20000Hz and the values are normalized to 8-bits.  Now your software can "watch" the 20-200Hz band for a signal above a threshold that is present for "X" mS.  After that time, you assume that there is a beat.  Then you wait for it to go away for "X" mS.  Now you assume that the beat is over.
I might have been inclined to look for peaks in the residuals from the Linear Prediction Coefficients
But windowed pulses in low FFT bands sounds lika promising idea too.
OdieAuthor Commented:

I'm nearly ready with implementing your method.


Could you explain that Linear Prediction Coefficient stuff a little more.  I might want to implement different methods to make sure I'm counting beats (and not noise or something).

Question has a verified solution.

Are you are experiencing a similar issue? Get a personalized answer when you ask a related question.

Have a better answer? Share it in a comment.

All Courses

From novice to tech pro — start learning today.