Fork Code: Spectrogram of a sound file

name

code content

// requires SignalBox quark

// 1. pick an fft size (must be power of two)
~size = 2048;
~size = 4096;

// 2. path to a sound file
~path = Platform.resourceDir +/+ "sounds/a11wlk01.wav";

(
// 3. analyze audio : this may take a little while for a long sound file
var data = Signal.read(~path);
var size = ~size;

var sf = SoundFile.openRead(~path);
var sampleRate = sf.sampleRate;

var hops = 4;
var window = Signal.hanningWindow(size);

var numFrames = (data.size / size * hops);
var fftMags = Array(numFrames);

var imag = Signal.newClear(size);
var cosTable = Signal.fftCosTable(size);

sf.close;

(data.size / size * hops).asInteger.do { |j|
  var i = j / hops;
  var r = data[(size * i).asInteger..(size * (i + 1) - 1).asInteger];
  r = r.addAll(Array.fill(size - r.size, {0}));
  r = r * window;
  fftMags.add((fft(r, imag, cosTable).magnitude.log10)[0..(size * 20000/sampleRate).asInteger]);
};

~fftMags = fftMags;
)

(
// 4. generate image : this may also take a little while
var width = ~fftMags.size;
var height = ~fftMags[0].size;
var image = Image.new(width, height);
var scaleFunc = { |pixel| pixel.lincurve(0, height, 1, height, curve: 3) }; // curve factor for Y scaleing (0 = linear, 3 =~ mel scale)
image.setPixels(Int32Array.fill(width * height, { |i|
  var index = scaleFunc.(height - (i / width) - 1).asInteger;
  var frame = i % width;
  var mag = ~fftMags[frame][index];
  Image.colorToPixel(Color.gray(0, mag.lincurve(-2, 1, 0, 1, curve: 1.5))); // input range to adjust threshold, curve factor for contrast: 0-2
}), Rect(0, 0, image.width, image.height));
~image.free;
~image = image;
)

// 5. display image
~image.plot;

code description use markdown for formating

category tags comma separated, i.g. "wild, siren" (do not enter default SC class names, please)

ancestor(s) comma separated identificators, i.g. "1-C,1-1,1-4M,1-x"

Private? the code will be accessible by direct url and not visible in public activity

comment of change