Gammatone Single gammatone filter 

Gammatone.ar(input, centrefrequency, bandwidth, mul, add )

A bandpass filter, as used in auditory modeling; approximates the frequency selectivity of a haircell site on the basilar membrane in the cochlea. 

input - input sound to filter.

centrefrequency - Centre frequency. Default value is 440.0. Cannot be modulated.

bandwidth - Bandwidth of filter. Default value is 200.0. Cannot be modulated.

See also: Meddis, HairCell

// what this example does

{ Gammatone.ar(WhiteNoise.ar) }.play;

//test sweeping frequency of sine input

{ Gammatone.ar(SinOsc.ar(MouseX.kr(100,2000)),1000.0,500.0) }.play;

//create gammatone filterbank with filters centred on 88 piano keys

(

var outereargain, freqs, bandwidths; 

freqs= Array.fill(88, {|i| (21+i).midicps});

//bandwidths in Hz not angular frequency for calculations in Hohmann

bandwidths= 24.7*(freqs*0.00437 +1);

//could reduce these bandwidths, ie 0.125*bandwidths, as test of resonance 

//calculated from ISO2002 100dB phon curve as approximation to sensitivity of outer ear

outereargain = [ 0.080993772688853, 0.089237063418603, 0.097970525722848, 0.10876266808815, 0.12105360022674, 0.13407538922284, 0.14787149408674, 0.16167718172822, 0.17551213214451, 0.19016975152285, 0.20569895830538, 0.22220242493157, 0.2397154127401, 0.2582697769952, 0.27792744116278, 0.29856170337058, 0.32031587344117, 0.34336361377938, 0.36778184407609, 0.39203618886184, 0.41615498938524, 0.44170796842, 0.46878040665827, 0.49446634124901, 0.52001934250193, 0.54709180427952, 0.57577407840633, 0.60165721569576, 0.62782392863276, 0.6555465952902, 0.68491773749099, 0.72411898089945, 0.77354432147194, 0.82590864573464, 0.88138671475204, 0.90157113760596, 0.90157113760596, 0.90157113760596, 0.90157113760596, 0.91510438365248, 0.93321426408541, 0.95240101404735, 0.97272866753273, 0.98926474289463, 1.0053814152966, 1.0224564349105, 1.0405467880267, 1.052760426677, 1.061847894439, 1.071475731154, 1.0816760688321, 1.0839269140212, 1.0839269140212, 1.0839269140212, 1.0839269140212, 1.0817917022137, 1.0789305551391, 1.0758992754061, 1.0726877464002, 1.0605734918597, 1.0429115831426, 1.0241994426808, 1.0043746204451, 0.95672713285027, 0.89882037142901, 0.83747029478932, 0.77247215275345, 0.75895909569467, 0.74934344623004, 0.73915602049396, 0.72836281890006, 0.75972195442662, 0.81665861045357, 0.87698089623035, 0.94089013177825, 1.022563004048, 1.1129841925092, 1.2087821046319, 1.3102764570426, 1.3485952672556, 1.3833144995125, 1.4200982447533, 1.4590692653081, 1.4180204618623, 1.3674299412878, 1.3138311518147, 1.2570452124657, 1.1677392221542 ];

//OK, make filterbank with 88 channels! mixed down for clarity...

{Mix(Gammatone.ar(WhiteNoise.ar,freqs, bandwidths)*outereargain)}.play;

//{Mix(Gammatone.ar(SoundIn.ar,freqs, bandwidths)*outereargain)}.play;

)

//20 channels with ERB scale spacing of centre frequencies. With each filter channel put through Meddis haircell models, may simulate what a cochlear implant 'sounds' like

(

var outereargain, freqs, bandwidths; 

var lowerbs, higherbs; 

var lowfreq, highfreq, numbands;

var prop, erbsnow; 

//create bank spaced equally in ERB scale units from lower frequency to higher

lowfreq = 100.0;

highfreq= 8000.0; //can increase this, maybe up to 16000.0?

numbands= 20; //try increasing this, nice around 50  

lowerbs = log10((lowfreq*0.00437 +1))*21.4; 

higherbs = log10((highfreq*0.00437 +1))*21.4; 

freqs= Array.fill(numbands,{arg i;   

prop= i.asFloat/(numbands-1);

erbsnow= lowerbs+((prop)*(higherbs-lowerbs)); 

228.8455*(pow(10, 0.046728972*erbsnow)-1);

});

bandwidths= 24.7*(freqs*0.00437 +1);

//OK, make filterbank with 20 channels, passing each into Meddis model mixed down for clarity...

{Mix(Meddis.ar(Gammatone.ar(SoundIn.ar,freqs, bandwidths)))}.play;

)

coded following: 

V Hohmann Frequency analysis and synthesis using a Gammatone filterbank Acta Acustica vol 88 (2002): 433--442