Help/UGens/MachineListening/KeyTrack.html

   1 <!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01//EN" "http://www.w3.org/TR/html4/strict.dtd">
   2 <html>
   3 <head>
   4 <meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
   5 <meta http-equiv="Content-Style-Type" content="text/css">
   6 <title></title>
   7 <meta name="Generator" content="Cocoa HTML Writer">
   8 <meta name="CocoaVersion" content="824.44">
   9 <style type="text/css">
  10 p.p1 {margin: 0.0px 0.0px 0.0px 0.0px; font: 18.0px Helvetica}
  11 p.p2 {margin: 0.0px 0.0px 0.0px 0.0px; font: 12.0px Helvetica; min-height: 14.0px}
  12 p.p3 {margin: 0.0px 0.0px 0.0px 0.0px; font: 12.0px Helvetica}
  13 p.p4 {margin: 0.0px 0.0px 0.0px 0.0px; font: 9.0px Monaco; color: #7c2a1b}
  14 p.p5 {margin: 0.0px 0.0px 0.0px 0.0px; font: 9.0px Monaco; min-height: 12.0px}
  15 p.p6 {margin: 0.0px 0.0px 0.0px 0.0px; font: 9.0px Monaco; color: #606060}
  16 p.p7 {margin: 0.0px 0.0px 0.0px 0.0px; font: 9.0px Monaco; color: #5b2e22}
  17 p.p8 {margin: 0.0px 0.0px 0.0px 0.0px; font: 9.0px Monaco}
  18 p.p9 {margin: 0.0px 0.0px 0.0px 0.0px; font: 9.0px Monaco; color: #872718}
  19 span.s1 {color: #000000}
  20 span.s2 {color: #1431aa}
  21 span.s3 {color: #25449a}
  22 span.s4 {color: #20409d}
  23 span.s5 {color: #102aae}
  24 span.s6 {color: #7c2a1b}
  25 span.Apple-tab-span {white-space:pre}
  26 </style>
  27 </head>
  28 <body>
  29 <p class="p1"><b>KeyTrack<span class="Apple-tab-span">  </span><span class="Apple-tab-span">    </span><span class="Apple-tab-span">    </span><span class="Apple-tab-span">    </span><span class="Apple-tab-span">    </span>key tracker</b></p>
  30 <p class="p2"><br></p>
  31 <p class="p3"><b>key = KeyTrack.kr(chain, keydecay=2.0, chromadecay= 1.0)</b></p>
  32 <p class="p2"><br></p>
  33 <p class="p3"><b>chain- [fft] </b>Audio input to track. This must have been pre-analysed by a 4096 size FFT. No other FFT sizes are valid except as noted below. <span class="Apple-converted-space"> </span></p>
  34 <p class="p2"><br></p>
  35 <p class="p4">//With standard hop of half FFT size = 2048 samples</p>
  36 <p class="p4"><span class="s1">b = </span><span class="s2">Buffer</span><span class="s1">.alloc(s,4096,1); </span>//for sampling rates 44100 and 48000</p>
  37 <p class="p4">//b = Buffer.alloc(s,8192,1); //for sampling rates 88200 and 96000</p>
  38 <p class="p2"><br></p>
  39 <p class="p3"><b>keydecay- [sk] </b>Number of seconds for the influence of a window on the final key decision to decay by 40dB (to 0.01 its original value)</p>
  40 <p class="p3"><b>chromaleak- [sk] </b>Each frame, the chroma values are set to the previous value multiplied by the chromadecay. 0.0 will start each frame afresh with no memory.<span class="Apple-converted-space"> </span></p>
  41 <p class="p2"><br></p>
  42 <p class="p2"><br></p>
  43 <p class="p3">A (12TET major/minor) key tracker based on a pitch class profile of energy across FFT bins and matching this to templates for major and minor scales in all transpositions. It assumes a 440 Hz concert A reference. Output is 0-11<span class="Apple-converted-space">  </span>C major to B major, 12-23 C minor to B minor</p>
  44 <p class="p5"><br></p>
  45 <p class="p5"><br></p>
  46 <p class="p5"><br></p>
  47 <p class="p5"><br></p>
  48 <p class="p5"><br></p>
  49 <p class="p4">//The following files are test materials on my machine; you will subsitute your own filenames here</p>
  50 <p class="p4">//A major</p>
  51 <p class="p6"><span class="s1">d=</span><span class="s2">Buffer</span><span class="s1">.read(s,</span>"/Volumes/data/stevebeattrack/samples/100.wav"<span class="s1">);</span></p>
  52 <p class="p4">//F major; hard to track!</p>
  53 <p class="p6"><span class="s1">d=</span><span class="s2">Buffer</span><span class="s1">.read(s,</span>"/Volumes/data/stevebeattrack/samples/115.wav"<span class="s1">);</span></p>
  54 <p class="p5"><br></p>
  55 <p class="p4">//straight forward since no transients; training set from MIREX2006<span class="Apple-converted-space"> </span></p>
  56 <p class="p4">//01= A major</p>
  57 <p class="p4">//57 = b minor</p>
  58 <p class="p4">//78 e minor</p>
  59 <p class="p4">//08 Bb major</p>
  60 <p class="p6"><span class="s1">d=</span><span class="s2">Buffer</span><span class="s1">.read(s,</span>"/Users/nickcollins/Desktop/ML/training_wav/78.wav"<span class="s1">)</span></p>
  61 <p class="p5"><br></p>
  62 <p class="p5"><br></p>
  63 <p class="p7"><span class="s1">b = </span><span class="s3">Buffer</span><span class="s1">.alloc(s,4096,1); </span>//for sampling rates 44100 and 48000</p>
  64 <p class="p5"><br></p>
  65 <p class="p5"><br></p>
  66 <p class="p8">(</p>
  67 <p class="p8">{</p>
  68 <p class="p5"><br></p>
  69 <p class="p8"><span class="s4">var</span> in, fft, resample;<span class="Apple-converted-space"> </span></p>
  70 <p class="p8"><span class="s4">var</span> key, transientdetection;</p>
  71 <p class="p5"><br></p>
  72 <p class="p8">in= <span class="s4">PlayBuf</span>.ar(1,d,<span class="s4">BufRateScale</span>.kr(d),1,0,1);</p>
  73 <p class="p5"><br></p>
  74 <p class="p8">fft = <span class="s4">FFT</span>(b, in);</p>
  75 <p class="p5"><br></p>
  76 <p class="p8">key=<span class="s4">KeyTrack</span>.kr(fft, 2.0, 0.5);</p>
  77 <p class="p5"><br></p>
  78 <p class="p8">key.poll;<span class="Apple-converted-space"> </span></p>
  79 <p class="p5"><br></p>
  80 <p class="p8"><span class="s4">Out</span>.ar(0,<span class="s4">Pan2</span>.ar(in));<span class="Apple-converted-space"> </span></p>
  81 <p class="p8">}.play</p>
  82 <p class="p8">)</p>
  83 <p class="p5"><br></p>
  84 <p class="p5"><br></p>
  85 <p class="p5"><br></p>
  86 <p class="p5"><br></p>
  87 <p class="p4">//alternating major and minor chords as a test</p>
  88 <p class="p8">(</p>
  89 <p class="p8">{</p>
  90 <p class="p5"><br></p>
  91 <p class="p8"><span class="s2">var</span> in, fft, resample;<span class="Apple-converted-space"> </span></p>
  92 <p class="p8"><span class="s2">var</span> key, transientdetection;</p>
  93 <p class="p5"><br></p>
  94 <p class="p8">in= <span class="s2">Mix</span>(<span class="s2">SinOsc</span>.ar((60+[0,<span class="s2">MouseX</span>.kr(3,4).round(1),7]).midicps,0,0.1));<span class="Apple-converted-space"> </span></p>
  95 <p class="p5"><br></p>
  96 <p class="p4">//major dom 7 and minor 7; major keys preferred here</p>
  97 <p class="p4">//in= Mix(SinOsc.ar((60+(MouseY.kr(0,11).round(1.0))+[0,MouseX.kr(3,4).round(1),7, 10]).midicps,0,0.1));<span class="Apple-converted-space"> </span></p>
  98 <p class="p5"><br></p>
  99 <p class="p8">fft = <span class="s2">FFT</span>(b, in);</p>
 100 <p class="p5"><br></p>
 101 <p class="p8">key=<span class="s2">KeyTrack</span>.kr(fft);</p>
 102 <p class="p5"><br></p>
 103 <p class="p8">key.poll;<span class="Apple-converted-space"> </span></p>
 104 <p class="p5"><br></p>
 105 <p class="p8"><span class="s2">Out</span>.ar(0,<span class="s2">Pan2</span>.ar(in));<span class="Apple-converted-space"> </span></p>
 106 <p class="p8">}.play</p>
 107 <p class="p8">)</p>
 108 <p class="p5"><br></p>
 109 <p class="p5"><br></p>
 110 <p class="p5"><br></p>
 111 <p class="p5"><br></p>
 112 <p class="p5"><br></p>
 113 <p class="p5"><br></p>
 114 <p class="p5"><br></p>
 115 <p class="p9">//Nice to hear what KeyTrack thinks:</p>
 116 <p class="p5"><br></p>
 117 <p class="p6"><span class="s1">d=</span><span class="s5">Buffer</span><span class="s1">.read(s,</span>"/Users/nickcollins/Desktop/ML/training_wav/78.wav"<span class="s1">)</span></p>
 118 <p class="p9"><span class="s1">b = </span><span class="s5">Buffer</span><span class="s1">.alloc(s,4096,1); </span>//for sampling rates 44100 and 48000</p>
 119 <p class="p5"><br></p>
 120 <p class="p8">(</p>
 121 <p class="p8">{</p>
 122 <p class="p5"><br></p>
 123 <p class="p8"><span class="s5">var</span> in, fft, resample, chord, rootnote, sympath;</p>
 124 <p class="p8"><span class="s5">var</span> key, transientdetection;</p>
 125 <p class="p5"><br></p>
 126 <p class="p8">in= <span class="s5">PlayBuf</span>.ar(1,d,<span class="s5">BufRateScale</span>.kr(d),1,0,1);</p>
 127 <p class="p5"><br></p>
 128 <p class="p8">fft = <span class="s5">FFT</span>(b, in);</p>
 129 <p class="p5"><br></p>
 130 <p class="p8">key=<span class="s5">KeyTrack</span>.kr(fft, 2.0, 0.5);</p>
 131 <p class="p8">key.poll;</p>
 132 <p class="p9"><span class="s1">key = </span><span class="s5">Median</span><span class="s1">.kr(101, key); </span>// Remove outlier wibbles</p>
 133 <p class="p5"><br></p>
 134 <p class="p8">chord = if(key&lt;12, #[0,4,7], #[0,3,7]);</p>
 135 <p class="p8">rootnote = if(key&lt;12, key, key-12) + 60;</p>
 136 <p class="p5"><br></p>
 137 <p class="p8">sympath = <span class="s5">SinOsc</span>.ar((rootnote + chord).midicps, 0, 0.4).mean;</p>
 138 <p class="p5"><br></p>
 139 <p class="p8"><span class="s5">Out</span>.ar(0,<span class="s5">Pan2</span>.ar(in, -0.5) + <span class="s5">Pan2</span>.ar(sympath, 0.5));</p>
 140 <p class="p8">}.play</p>
 141 <p class="p8">)</p>
 142 <p class="p5"><br></p>
 143 <p class="p5"><br></p>
 144 <p class="p5"><br></p>
 145 <p class="p5"><br></p>
 146 <p class="p5"><br></p>
 147 <p class="p5"><br></p>
 148 <p class="p5"><br></p>
 149 <p class="p5"><br></p>
 150 <p class="p5"><br></p>
 151 <p class="p5"><br></p>
 152 <p class="p4">//Research Notes:<span class="Apple-converted-space"> </span></p>
 153 <p class="p4">//See the MIREX2006 audio key tracking competition and Emilia Gomez's PhD thesis, Tonal Description of Music Audio Signals</p>
 154 <p class="p5"><br></p>
 155 <p class="p4">//The following code was used to create the datasets for the UGen, and would be the basis of extensions<span class="Apple-converted-space"> </span></p>
 156 <p class="p5"><br></p>
 157 <p class="p4">//Need one set of bin data for 44100 and one for 48000</p>
 158 <p class="p5"><br></p>
 159 <p class="p4">//KeyTrack calculations, need to make arrays of FFT bins and weights for each chromatic tone.<span class="Apple-converted-space"> </span></p>
 160 <p class="p4">//greater resolution, 4096 FFT, avoid lower octaves, too messy there</p>
 161 <p class="p4">//60*6*2 output arrays</p>
 162 <p class="p5"><br></p>
 163 <p class="p8">(</p>
 164 <p class="p8"><span class="s2">var</span> fftN, fftBins, binsize;</p>
 165 <p class="p8"><span class="s2">var</span> midinotes;</p>
 166 <p class="p8"><span class="s2">var</span> sr;</p>
 167 <p class="p8"><span class="s2">var</span> wtlist, binlist;</p>
 168 <p class="p5"><br></p>
 169 <p class="p8">sr = 48000; <span class="s6">//44100;</span></p>
 170 <p class="p5"><br></p>
 171 <p class="p8">fftN=4096;<span class="Apple-converted-space"> </span></p>
 172 <p class="p8">fftBins=fftN.div(2);</p>
 173 <p class="p8">binsize=sr/fftN;</p>
 174 <p class="p5"><br></p>
 175 <p class="p4"><span class="s1">midinotes= (33..92); </span>//60 notes, 55 Hz up to 1661.2187903198 Hz</p>
 176 <p class="p5"><br></p>
 177 <p class="p8">wtlist=<span class="s2">List</span>[];<span class="Apple-converted-space"> </span></p>
 178 <p class="p8">binlist=<span class="s2">List</span>[];<span class="Apple-converted-space"> </span></p>
 179 <p class="p5"><span class="Apple-tab-span">     </span></p>
 180 <p class="p4">//for each note have six harmonic locations<span class="Apple-tab-span">  </span></p>
 181 <p class="p8">midinotes.do{<span class="s2">|note|</span></p>
 182 <p class="p8"><span class="Apple-tab-span">     </span><span class="s2">var</span> freq, whichbin, lowerbin, prop;<span class="Apple-converted-space"> </span></p>
 183 <p class="p5"><br></p>
 184 <p class="p8"><span class="Apple-tab-span">     </span>freq= note.midicps;<span class="Apple-converted-space"> </span></p>
 185 <p class="p5"><span class="Apple-tab-span">     </span></p>
 186 <p class="p8"><span class="Apple-tab-span">     </span>6.do{<span class="s2">|j|</span></p>
 187 <p class="p8"><span class="Apple-tab-span">     </span><span class="Apple-tab-span">    </span><span class="s2">var</span> partialfreq, partialamp;</p>
 188 <p class="p5"><span class="Apple-tab-span">     </span><span class="Apple-tab-span">    </span></p>
 189 <p class="p8"><span class="Apple-tab-span">     </span><span class="Apple-tab-span">    </span>partialamp= 1.0/(j+1);</p>
 190 <p class="p8"><span class="Apple-tab-span">     </span><span class="Apple-tab-span">    </span>partialfreq= freq*(j+1);<span class="Apple-converted-space"> </span></p>
 191 <p class="p5"><span class="Apple-tab-span">     </span><span class="Apple-tab-span">    </span></p>
 192 <p class="p8"><span class="Apple-tab-span">     </span><span class="Apple-tab-span">    </span>whichbin= partialfreq/binsize;<span class="Apple-converted-space"> </span></p>
 193 <p class="p8"><span class="Apple-tab-span">     </span><span class="Apple-tab-span">    </span>lowerbin= whichbin.asInteger;<span class="Apple-converted-space"> </span></p>
 194 <p class="p8"><span class="Apple-tab-span">     </span><span class="Apple-tab-span">    </span>prop= 1.0-(whichbin-lowerbin);</p>
 195 <p class="p5"><span class="Apple-tab-span">     </span><span class="Apple-tab-span">    </span></p>
 196 <p class="p8"><span class="Apple-tab-span">     </span><span class="Apple-tab-span">    </span>binlist.add(lowerbin).add(lowerbin+1);</p>
 197 <p class="p8"><span class="Apple-tab-span">     </span><span class="Apple-tab-span">    </span>wtlist.add(prop*partialamp).add((1.0-prop)*partialamp);</p>
 198 <p class="p5"><span class="Apple-tab-span">     </span></p>
 199 <p class="p8"><span class="Apple-tab-span">     </span>};</p>
 200 <p class="p5"><br></p>
 201 <p class="p8">};</p>
 202 <p class="p5"><br></p>
 203 <p class="p5"><br></p>
 204 <p class="p8"><span class="s2">Post</span> &lt;&lt; (binlist) &lt;&lt; nl&lt;&lt; nl;</p>
 205 <p class="p5"><br></p>
 206 <p class="p8"><span class="s2">Post</span> &lt;&lt; (wtlist) &lt;&lt; nl&lt;&lt; nl;</p>
 207 <p class="p5"><br></p>
 208 <p class="p8">binlist.size.postln;</p>
 209 <p class="p8">wtlist.size.postln;</p>
 210 <p class="p8">)</p>
 211 <p class="p5"><br></p>
 212 </body>
 213 </html>