MAT240E-Final II [Raag Analysis]

gistfile1.txt

{
 "metadata": {
  "name": ""
 },
 "nbformat": 3,
 "nbformat_minor": 0,
 "worksheets": [
  {
   "cells": [
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "#Muhammad Hafiz Wan Rosli\n",
      "#MAT240E-Final II [Raag Analysis]\n",
      "#Raag Analysis\n",
      "#Source: Master of Sitar- 01 Raag Alahya Bilawal (Early Morning Raag).mp3\n",
      "#Ground Truth:\n",
      "#Jhala: 00:00 - 00:42\n",
      "#Gat:   00:42 - 18:46 (1080+46 = 1126 secs)\n",
      "#Jhala: 18:46 - end (check spectral difference 1 minute before & after)\n",
      "#Gat-Jhala transition: 18:45 - 18:46\n",
      "\n",
      "#Code written in python        : https://www.python.org/\n",
      "#Interactive computing ipython : http://ipython.org/\n",
      "#Audio Analysis & MIR Library  : http://essentia.upf.edu/"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 1
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "import numpy as np\n",
      "from essentia import *\n",
      "from essentia.streaming import *\n",
      "import matplotlib.pyplot as plt\n",
      "from matplotlib.widgets import Slider, Button, RadioButtons, CheckButtons\n",
      "plt.rcParams['figure.figsize'] = (16,9)"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 2
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "sr = 44100\n",
      "frameSize = 2048\n",
      "hopSize = 1024\n",
      "\n",
      "loader = MonoLoader(filename = '/Users/muhammadhafiz/Documents/work/ComputationalEthno/samples/RaagAlahyaBilawal0_3minutes.mp3', sampleRate=sr)\n",
      "frameCutter = FrameCutter(frameSize = frameSize, hopSize = hopSize)\n",
      "w = Windowing(type = 'hann')\n",
      "pitchyin = PitchYinFFT()\n",
      "spec = Spectrum()\n",
      "mfcc = MFCC()\n",
      "centroid = Centroid()\n",
      "rolloff = RollOff()\n",
      "flux = Flux()\n",
      "zcr = ZeroCrossingRate()\n",
      "rms = RMS()\n",
      "\n",
      "#initialize algorithms\n",
      "onsetDetector = OnsetRate()\n",
      "beatsLoudness = BeatsLoudness()\n",
      "centralMoments = CentralMoments()\n",
      "derivativeSFX = DerivativeSFX()\n",
      "envelope = Envelope()\n",
      "#distributionShape = DistributionShape()\n",
      "erbBands = ERBBands()\n",
      "frequencyBands = FrequencyBands()#frequencyBands([141, 142, 213, 282, 284, 286, 355, 358, 428]))\n",
      "#harmonicPeaks = HarmonicPeaks()\n",
      "#oddToEvenHarmonicEnergyRatio = OddToEvenHarmonicEnergyRatio()\n",
      "#pitchContours = PitchContours()\n",
      "#pitchContoursMelody = PitchCountoursMelody()\n",
      "pitchSalience = PitchSalience()\n",
      "pitchSalienceFunction = PitchSalienceFunction()###\n",
      "pitchSalienceFunctionPeaks = PitchSalienceFunctionPeaks()###\n",
      "predominantMelody = PredominantMelody()\n",
      "SBIC = SBic()#\n",
      "spectralComplexity = SpectralComplexity()\n",
      "spectralContrast = SpectralContrast()\n",
      "spectralPeaks = SpectralPeaks()\n",
      "#tcToTotal = TCToTotal()"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 3
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "#connect sink+source networks\n",
      "loader.audio >> onsetDetector.signal#\n",
      "loader.audio >> beatsLoudness.signal\n",
      "#loader.audio >> envelope.signal \n",
      "loader.audio >> predominantMelody.signal\n",
      "loader.audio >> frameCutter.signal\n",
      "frameCutter.frame >> w.frame >> spec.frame\n",
      "frameCutter.frame >> zcr.signal\n",
      "frameCutter.frame >> rms.array\n",
      "#w.frame >> centralMoments.array\n",
      "#envelope.signal >> derivativeSFX.envelope\n",
      "\n",
      "spec.spectrum >> centroid.array\n",
      "spec.spectrum >> erbBands.spectrum\n",
      "spec.spectrum >> flux.spectrum\n",
      "spec.spectrum >> frequencyBands.spectrum\n",
      "spec.spectrum >> mfcc.spectrum\n",
      "spec.spectrum >> pitchSalience.spectrum\n",
      "spec.spectrum >> pitchyin.spectrum\n",
      "spec.spectrum >> rolloff.spectrum\n",
      "spec.spectrum >> spectralComplexity.spectrum\n",
      "spec.spectrum >> spectralContrast.spectrum###\n",
      "spec.spectrum >> spectralPeaks.spectrum\n",
      "\n",
      "#spectralPeaks.frequencies >> pitchSalienceFunction.frequencies\n",
      "#spectralPeaks.magnitudes >> pitchSalienceFunction.magnitudes\n",
      "#pitchSalienceFunction.salienceFunction >> pitchSalienceFunctionPeaks.salienceFunction"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 4,
       "text": [
        "<essentia.streaming._StreamConnector instance at 0x11141ba70>"
       ]
      }
     ],
     "prompt_number": 4
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "#connect to individual pools\n",
      "pool = essentia.Pool()\n",
      "\n",
      "onsetDetector.onsetTimes >> (pool, 'onset.onsetTimes')\n",
      "onsetDetector.onsetRate >> (pool, 'onset.onsetRate')\n",
      "beatsLoudness.loudness >> (pool, 'beatsLoudness.loudness')\n",
      "beatsLoudness.loudnessBandRatio >> (pool, 'beatsLoudness.loudnessBandRatio')\n",
      "#derivativeSFX.derAvAfterMax >> (pool, 'derivativeSFX.derAvAfterMax')\n",
      "#derivativeSFX.maxDerBeforeMax >> (pool, 'derivativeSFX.maxDerBeforeMax')\n",
      "erbBands.bands >> (pool, 'erbBands.bands')\n",
      "frequencyBands.bands >> (pool, 'frequencyBands.bands')\n",
      "pitchSalience.pitchSalience >> (pool, 'pitchSalience.pitchSalience')\n",
      "#pitchSalienceFunction.salienceFunction >> (pool, 'pitchSalienceFunction.salienceFunction')\n",
      "#pitchSalienceFunctionPeaks.salienceBins >> (pool, 'pitchSalienceFunctionPeaks.salienceBins')\n",
      "#pitchSalienceFunctionPeaks.salienceValues >> (pool, 'pitchSalienceFunctionPeaks.salienceValues')\n",
      "predominantMelody.pitch >> (pool, 'predominantMelody.pitch')\n",
      "predominantMelody.pitchConfidence >> (pool, 'predominantMelody.pitchConfidence')\n",
      "spectralComplexity.spectralComplexity >> (pool, 'spectralComplexity.spectralComplexity')\n",
      "spectralContrast.spectralContrast >> (pool, 'spectralContrast.spectralContrast')###\n",
      "spectralContrast.spectralValley >> (pool, 'spectralContrast.spectralValley')###\n",
      "spectralPeaks.frequencies >> (pool, 'spectralPeaks.frequencies')\n",
      "spectralPeaks.magnitudes >> (pool, 'spectralPeaks.magnitudes')\n",
      "\n",
      "\n",
      "pitchyin.pitch >> (pool, 'pitchyin.pitch')\n",
      "pitchyin.pitchConfidence >> (pool, 'pitchyin.pitchConfidence')\n",
      "spec.spectrum >> (pool, 'lowlevel.spectrum')\n",
      "mfcc.bands >> (pool, 'lowlevel.mfcc_bands')\n",
      "mfcc.mfcc >> (pool, 'lowlevel.mfcc')\n",
      "centroid.centroid >> (pool, 'lowlevel.centroid')\n",
      "rolloff.rollOff >> (pool, 'lowlevel.rolloff')\n",
      "flux.flux >> (pool, 'lowlevel.flux')\n",
      "\n",
      "zcr.zeroCrossingRate >> (pool, 'lowlevel.zcr')\n",
      "rms.rms >> (pool, 'lowlevel.rms')"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 5
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "essentia.reset(loader)"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 6
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "essentia.run(loader)"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 7
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "timeConversion = 44100/ float(hopSize)\n",
      "groundTruthAlap=42 #secs\n",
      "analysisLength = 60\n",
      "\n",
      "#Normalize pools 0-100 (percent)\n",
      "pitchYinPool           = (pool['pitchyin.pitch'][:analysisLength*timeConversion]                           / max(pool['pitchyin.pitch'][:analysisLength*timeConversion])) * 100\n",
      "pitchYinConfidencePool = (pool['pitchyin.pitchConfidence'][:analysisLength*timeConversion]                 / max(pool['pitchyin.pitchConfidence'][:analysisLength*timeConversion])) * 100\n",
      "centroidPool           = (pool['lowlevel.centroid'][:analysisLength*timeConversion]                        / max(pool['lowlevel.centroid'][:analysisLength*timeConversion])) * 100\n",
      "rolloffPool            = (pool['lowlevel.rolloff'][:analysisLength*timeConversion]                         / max(pool['lowlevel.rolloff'][:analysisLength*timeConversion])) * 100\n",
      "fluxPool               = (pool['lowlevel.flux'][:analysisLength*timeConversion]                            / max(pool['lowlevel.flux'][:analysisLength*timeConversion])) * 100\n",
      "zcrPool                = (pool['lowlevel.zcr'][:analysisLength*timeConversion]                             / max(pool['lowlevel.zcr'][:analysisLength*timeConversion])) * 100\n",
      "rmsPool                = (pool['lowlevel.rms'][:analysisLength*timeConversion]                             / max(pool['lowlevel.rms'][:analysisLength*timeConversion])) * 100\n",
      "pitchSaliencePool      = (pool['pitchSalience.pitchSalience'][:analysisLength*timeConversion]              / max(pool['pitchSalience.pitchSalience'][:analysisLength*timeConversion])) * 100\n",
      "spectralComplexityPool = (pool['spectralComplexity.spectralComplexity'][:analysisLength*timeConversion]    / max(pool['spectralComplexity.spectralComplexity'][:analysisLength*timeConversion])) * 100\n",
      "\n",
      "#plot values\n",
      "fig, ax1 = plt.subplots()\n",
      "l = ax1.vlines(groundTruthAlap, 0, 100, lw=3, alpha = 0.5, color='k')\n",
      "l0, = ax1.plot(np.linspace(0, analysisLength, len(pool['pitchyin.pitch'][:analysisLength*timeConversion])), pitchYinPool, visible=False)\n",
      "l1, = ax1.plot(np.linspace(0, analysisLength, len(pool['pitchyin.pitchConfidence'][:analysisLength*timeConversion])), pitchYinConfidencePool, visible=False)\n",
      "l2, = ax1.plot(np.linspace(0, analysisLength, len(pool['lowlevel.centroid'][:analysisLength*timeConversion])), centroidPool, visible=False)\n",
      "p2  = ax1.vlines((np.argmax(pool['lowlevel.centroid'][10*timeConversion: analysisLength*timeConversion])/ timeConversion)+ 10, 0, (max(pool['lowlevel.centroid'][10*timeConversion: analysisLength*timeConversion])/ max(pool['lowlevel.centroid'][:analysisLength*timeConversion]))*100, color ='r', visible=False)\n",
      "l3, = ax1.plot(np.linspace(0, analysisLength, len(pool['lowlevel.rolloff'][:analysisLength*timeConversion])), rolloffPool, visible=False)\n",
      "p3  = ax1.vlines((np.argmax(pool['lowlevel.rolloff'][10*timeConversion: analysisLength*timeConversion])/ timeConversion)+ 10, 0, (max(pool['lowlevel.rolloff'][10*timeConversion: analysisLength*timeConversion])/ max(pool['lowlevel.rolloff'][:analysisLength*timeConversion]))*100, color ='r', visible=False)\n",
      "l4, = ax1.plot(np.linspace(0, analysisLength, len(pool['lowlevel.flux'][:analysisLength*timeConversion])), fluxPool, visible=False)\n",
      "p4  = ax1.vlines((np.argmax(pool['lowlevel.flux'][10*timeConversion: analysisLength*timeConversion])/ timeConversion)+ 10, 0, (max(pool['lowlevel.flux'][10*timeConversion: analysisLength*timeConversion])/ max(pool['lowlevel.flux'][:analysisLength*timeConversion]))*100, color ='r', visible=False)\n",
      "l5, = ax1.plot(np.linspace(0, analysisLength, len(pool['lowlevel.zcr'][:analysisLength*timeConversion])), zcrPool, visible=False)\n",
      "p5  = ax1.vlines((np.argmax(pool['lowlevel.zcr'][10*timeConversion: analysisLength*timeConversion])/ timeConversion)+ 10, 0, (max(pool['lowlevel.zcr'][10*timeConversion: analysisLength*timeConversion])/ max(pool['lowlevel.zcr'][:analysisLength*timeConversion]))*100, color ='r', visible=False)\n",
      "l6, = ax1.plot(np.linspace(0, analysisLength, len(pool['lowlevel.rms'][:analysisLength*timeConversion])), rmsPool, visible=False)\n",
      "p6  = ax1.vlines((np.argmax(pool['lowlevel.rms'][10*timeConversion: analysisLength*timeConversion])/ timeConversion)+ 10, 0, (max(pool['lowlevel.rms'][10*timeConversion: analysisLength*timeConversion])/ max(pool['lowlevel.rms'][:analysisLength*timeConversion]))*100, color ='r', visible=False)\n",
      "l7, = ax1.plot(np.linspace(0, analysisLength, len(pool['pitchSalience.pitchSalience'][:analysisLength*timeConversion])), pitchSaliencePool, visible=False)\n",
      "p7  = ax1.vlines((np.argmax(pool['pitchSalience.pitchSalience'][10*timeConversion: analysisLength*timeConversion])/ timeConversion)+ 10, 0, (max(pool['pitchSalience.pitchSalience'][10*timeConversion: analysisLength*timeConversion])/ max(pool['pitchSalience.pitchSalience'][:analysisLength*timeConversion]))*100, color ='r', visible=False)\n",
      "l8, = ax1.plot(np.linspace(0, analysisLength, len(pool['spectralComplexity.spectralComplexity'][:analysisLength*timeConversion])), spectralComplexityPool, visible=False)\n",
      "p8  = ax1.vlines((np.argmax(pool['spectralComplexity.spectralComplexity'][10*timeConversion: analysisLength*timeConversion])/ timeConversion)+ 10, 0, (max(pool['spectralComplexity.spectralComplexity'][10*timeConversion: analysisLength*timeConversion])/ max(pool['spectralComplexity.spectralComplexity'][:analysisLength*timeConversion]))*100, color ='r', visible=False)\n",
      "#plot(linspace(0, 180, len(pool['erbBands.bands'][:180*timeConversion])), pool['erbBands.bands'][:180*timeConversion])\n",
      "\n",
      "\n",
      "\n",
      "#plt.legend(['pitch', 'pitch confidence', 'spectral centroid', 'spectral rolloff', 'spectral flux', 'ZCR', 'RMS', 'onset'])\n",
      "plt.suptitle('Raag Analysis: Alahya Bilawal [Ravi Shankar - Master of Sitar]', fontsize=14, fontweight='bold')\n",
      "plt.title('Ground truth- Alap: 42 seconds')\n",
      "plt.ylabel('percentage to maximum in analysis section')\n",
      "plt.xlabel('time (seconds)')\n",
      "plt.subplots_adjust(left=0.3)#, bottom=0.25)\n",
      "rax = plt.axes([0.025, 0.65, 0.22, 0.25])#left,bottom, width, height\n",
      "check = CheckButtons(rax, ('pitch', 'pitch confidence', 'spectral centroid', 'spectral rolloff', 'spectral flux', 'ZCR', 'RMS', 'pitch salience', 'spectral complexity'), (False, False, False, False, False, False, False, False, False))\n",
      "\n",
      "#checkbox controls\n",
      "def displayMode(label):\n",
      "    if label == 'pitch': \n",
      "        l0.set_visible(not l0.get_visible())\n",
      "    elif label == 'pitch confidence': \n",
      "        l1.set_visible(not l1.get_visible())\n",
      "    elif label == 'spectral centroid': \n",
      "        l2.set_visible(not l2.get_visible())\n",
      "        p2.set_visible(not p2.get_visible())\n",
      "    elif label == 'spectral rolloff': \n",
      "        l3.set_visible(not l3.get_visible())\n",
      "        p3.set_visible(not p3.get_visible())\n",
      "    elif label == 'spectral flux': \n",
      "        l4.set_visible(not l4.get_visible())\n",
      "        p4.set_visible(not p4.get_visible())\n",
      "    elif label == 'ZCR':\n",
      "        l5.set_visible(not l5.get_visible())\n",
      "        p5.set_visible(not p5.get_visible())\n",
      "    elif label == 'RMS': \n",
      "        l6.set_visible(not l6.get_visible())\n",
      "        p6.set_visible(not p6.get_visible())\n",
      "    elif label == 'pitch salience': \n",
      "        l7.set_visible(not l7.get_visible())\n",
      "        p7.set_visible(not p7.get_visible())\n",
      "    elif label == 'spectral complexity': \n",
      "        l8.set_visible(not l8.get_visible())\n",
      "        p8.set_visible(not p8.get_visible())\n",
      "    plt.draw()\n",
      "check.on_clicked(displayMode)\n",
      "plt.show()"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 8
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [],
     "language": "python",
     "metadata": {},
     "outputs": []
    }
   ],
   "metadata": {}
  }
 ]
}