A simple beat detection routine for openFrameworks that gives 32 fft bands plus separate output for kick, snare, and hihat. Created for inclusion in the Cyril language for live coded visuals.
This builds on the beat tracking code in ofxBeatTracking, which itself was based on C++ code from Dominic Mazzoni, which was based on an free FFT implementation in Fortan. Further details are in the comments here.
Install the addon as usual, then in your openFrameworks app define an instance of ofxBeat, for example:
class testApp : public ofBaseApp{
ofxBeat beat;
void audioReceived(float*, int, int);
// ... rest of your app class ...
};
Then initialize the sound stream in your app setup() method, for example:
void testApp::setup(){
ofSoundStreamSetup(0, 1, this, 44100, beat.getBufferSize(), 4);
}
Add the following method to pass on audio events to the FFT processor:
void testApp::audioReceived(float* input, int bufferSize, int nChannels) {
beat.audioReceived(input, bufferSize, nChannels);
}
Add a call to the update method to your app's update method to perform the fft processing on recevied audio, for example:
void testApp::update() {
beat.update(ofGetElapsedTimeMillis());
}
You can then use the separate bands to get information in the current beat graph. These methods return a value between 1 and 0. It's 1 when a beat is detected and fades away to 0 based on the detected time between beats.
float kick = beat.kick();
float snare = beat.snare();
float hihat = beat.hihat();
You can also access the raw fft bands using the following function, where i
is a value between 0 and 31 (i.e. 32 fft bands indexed from 0):
float selectedBand = beat.getBand(i);
This is a pretty basic implementation of beat detection but it is fast and gets a fairly good result.