*  Pitch Detection using Auto Correlation.

 *  

 *  Auto correlation multiplies each sample in a buffer by all

 *  of the other samples. This emphasizes the fundamental

 *  frequency.

 *

 *  Running the signal through a low pass filter prior to

 *  autocorrelation helps bring out the fundamental frequency.

 *  

 *  The visualization is a correlogram, which plots

 *  the autocorrelations.

 *

 *  We calculate the pitch by counting the number of samples

 *  between peaks.

 *  

 *  Example by Jason Sigal and Golan Levin.

 */

​

var source, fft, lowPass;

​

// center clip nullifies samples below a clip amount

var doCenterClip = true;

var centerClipThreshold = 20.0;

​

// normalize pre / post autocorrelation

var preNormalize = true;

var postNormalize = true;

​

​

function setup() {

  createCanvas(windowWidth, windowHeight);

  noFill();

    textSize(24);

    textAlign(CENTER, CENTER);

  textFont('monospace');

​

  source = new p5.AudioIn();

  source.start();

​

  lowPass = new p5.LowPass();

  lowPass.disconnect();

  source.connect(lowPass);

​

  fft = new p5.FFT();

  fft.setInput(lowPass);

}

​

function draw() {

  background(255);

  noFill();

​

  // array of values from -1 to 1

  var timeDomain = fft.waveform(1024, 'float32');

  var corrBuff = autoCorrelate(timeDomain);

​

  strokeWeight(3);

  stroke(66, 244, 155);

  beginShape();

  for (var i = 0; i < corrBuff.length; i++) {

    var w = map(i, 0, corrBuff.length, width/2, 0);

    var h = map(corrBuff[i], -1, 1, height, 0);

    curveVertex(w, h);

  }

  endShape();

    beginShape();

  for (var i = 0; i < corrBuff.length; i++) {

    var w = map(i, 0, corrBuff.length, width/2, width);

    var h = map(corrBuff[i], -1, 1, height, 0);

    curveVertex(w, h);

  }

  endShape();

​

  strokeWeight(3);

​

  stroke(20);

  line(0, height/2, width, height/2);

​

  var freq = findFrequency(corrBuff);

  fill(0);

    noStroke();

  text ('Fundamental Frequency: ' + (freq.toString() + '000').slice(0, 5) + 'Hz', width/2, height - 50); 

}

​

​

​

// accepts a timeDomainBuffer and multiplies every value

function autoCorrelate(timeDomainBuffer) {

  var nSamples = timeDomainBuffer.length;

​

  // pre-normalize the input buffer

  if (preNormalize){

    timeDomainBuffer = normalize(timeDomainBuffer);

  }

​

  // zero out any values below the centerClipThreshold

  if (doCenterClip) {

    timeDomainBuffer = centerClip(timeDomainBuffer);

  }

​

  var autoCorrBuffer = [];

  for (var lag = 0; lag < nSamples; lag++){

    var sum = 0; 

    for (var index = 0; index < nSamples-lag; index++){

      var indexLagged = index+lag;

      var sound1 = timeDomainBuffer[index];

      var sound2 = timeDomainBuffer[indexLagged];

      var product = sound1 * sound2;

      sum += product;

    }

​

    // average to a value between -1 and 1

    autoCorrBuffer[lag] = sum/nSamples;

  }

​

  // normalize the output buffer

  if (postNormalize){

    autoCorrBuffer = normalize(autoCorrBuffer);

  }

​

  return autoCorrBuffer;

}

​

​

// Find the biggest value in a buffer, set that value to 1.0,

// and scale every other value by the same amount.

function normalize(buffer) {

  var biggestVal = 0;

  var nSamples = buffer.length;

  for (var index = 0; index < nSamples; index++){

    if (abs(buffer[index]) > biggestVal){

      biggestVal = abs(buffer[index]);

    }

  }

  for (var index = 0; index < nSamples; index++){

​

    // divide each sample of the buffer by the biggest val

    buffer[index] /= biggestVal;

  }

  return buffer;

}

​

// Accepts a buffer of samples, and sets any samples whose

// amplitude is below the centerClipThreshold to zero.

// This factors them out of the autocorrelation.

function centerClip(buffer) {

  var nSamples = buffer.length;

​

  // center clip removes any samples whose abs is less than centerClipThreshold

  centerClipThreshold = map(mouseY, 0, height, 0,1); 

​

  if (centerClipThreshold > 0.0) {

    for (var i = 0; i < nSamples; i++) {

      var val = buffer[i];

      buffer[i] = (Math.abs(val) > centerClipThreshold) ? val : 0;

    }

  }

  return buffer;

}

​

// Calculate the fundamental frequency of a buffer

// by finding the peaks, and counting the distance

// between peaks in samples, and converting that

// number of samples to a frequency value.

function findFrequency(autocorr) {

​

  var nSamples = autocorr.length;

  var valOfLargestPeakSoFar = 0;

  var indexOfLargestPeakSoFar = -1;

​

  for (var index = 1; index < nSamples; index++){

    var valL = autocorr[index-1];

    var valC = autocorr[index];

    var valR = autocorr[index+1];

​

    var bIsPeak = ((valL < valC) && (valR < valC));

    if (bIsPeak){

      if (valC > valOfLargestPeakSoFar){

        valOfLargestPeakSoFar = valC;

        indexOfLargestPeakSoFar = index;

      }

    }

  }

  var distanceToNextLargestPeak = indexOfLargestPeakSoFar - 0;

​

  // convert sample count to frequency

  var fundamentalFrequency = sampleRate() / distanceToNextLargestPeak;

  return fundamentalFrequency;

}