jmbeach / beatdetector

A simple beat detector for visualising sound on the web. Also has limited BPM-detection support.

BeatDetector.js

---

A pretty rudimentary but working beat detector. Built using the Web audio api. Based on comparing average shift in freq amplitudes in a current sample to a sample history. Catches heavy beat hits pretty accurately (techno, house, hip hop, that kinda stuff)

BPM calculation funcionality mostly for kicks: don't use it for anything exact, there are far better options.

Based on the frequency select-algorithm (the web audio api does fft-calculations for us) but without band splitting from: http://archive.gamedev.net/archive/reference/programming/features/beatdetection/index.html

Example:

var song = new stasilo.BeatDetector({sens: 				5.0,
						 			 visualizerFFTSize: 256, 
									 analyserFFTSize:   256, 
									 passFreq:          600,
									 url:               "file.mp3" } ); 

Usage

To get frequency data for drawing bars and what not:

song.getAudioFreqData();

This returns an array of visualizerFFTSize / 2 data values corresponding to frequency amplitudes.

To detect a beat hit

Call

song.isOnBeat()

from the render loop of your script. Returns true if song is on a beat.

For everything else, see the source.

Settings

sens: Sensitivity of the algorithm. A value between 1 and 16 (1 - low threshold, 16 - high treshold) should do it. Requires a bit of trail and error tweaking for the sweet spot.

This setting is required.

url: Url to audio file.

Detection defaults to microphone if no url is supplied.

visualizerFFTSize: Size of fft calculations for visualizations. Must be a power of two (2^7 = 128, 2^8 = 256, 2^9 = 512, ...)

Default value: 256

analyserFFTSize: Size of fft calculation for the algorithm Must be a power of two (2^7 = 128, 2^8 = 256, 2^9 = 512, ...)

Default value: 256

passFreq: Float. If supplied, passes audio through a bandpass filter with a peak at this frequency before passing it on to the algorithm. Suitable for example when a song has a loud treble/mid section and you'd like to detect bass drum beats, in which case a bandpass at 100-800Hz could help you out.

Freq chart for common instruments: http://www.independentrecording.net/irn/resources/freqchart/main_display.htm

Default value: off

loop: Boolean. Whether to loop the sound or not.

Default: false.

playbackFinished: A function called at the end of playback.

progress(obj): A callback run while sound is downloading from url. An object of {percent: value, complete: boolean} is passed as an argument. Useful for when loading sounds through ajax.

Browser support

Please see: http://caniuse.com/#feat=audio-api

Contact

Jakob Stasilowicz made this. Contact me through kontakt [at] stasilo.se or http:///www.stasilo.se.