DTW_MFCC_KNN.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Load required libraries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import time\n",
    "import librosa\n",
    "from dtw import dtw\n",
    "import librosa.display\n",
    "from scipy.spatial.distance import cdist"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Load two sample audio data to memory"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "y1, sr1 = librosa.load('train/fcmc0-a1-t.wav')\n",
    "y2, sr2 = librosa.load('train/fcmc0-b1-t.wav')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Calculate MFCC features and plot on graph"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Populating the interactive namespace from numpy and matplotlib\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7f02ad8a0ef0>"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWQAAADuCAYAAAAOR30qAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nkpEAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f02afb4f978>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%pylab inline\n",
    "subplot(1, 2, 1)\n",
    "mfcc1 = librosa.feature.mfcc(y1, sr1)\n",
    "librosa.display.specshow(mfcc1)\n",
    "\n",
    "subplot(1, 2, 2)\n",
    "mfcc2 = librosa.feature.mfcc(y2, sr2)\n",
    "librosa.display.specshow(mfcc2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Normalized distance between the two sounds: 141.228637029\n"
     ]
    }
   ],
   "source": [
    "# Calculate the DTW between the 2 sample audios 'a' and 'b'\n",
    "dist, cost, path, _ = dtw(mfcc1.T, mfcc2.T, dist=lambda x, y: norm(x - y, ord=1))\n",
    "print('Normalized distance between the two sounds:', dist)   "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Main Program"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import os\n",
    "dirname = \"train\"\n",
    "files = [f for f in os.listdir(dirname) if not f.startswith('.')]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The following code Iterates through the Training folder and builds the trained representation in the Distance matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Time used: 160.74331600000002s\n"
     ]
    }
   ],
   "source": [
    "start = time.clock()\n",
    "minval = 200\n",
    "distances = ones((len(files), len(files)))\n",
    "y = ones(len(files))\n",
    "\n",
    "for i in range(len(files)):\n",
    "    y1, sr1 = librosa.load(dirname+\"/\"+files[i])\n",
    "    mfcc1 = librosa.feature.mfcc(y1, sr1)\n",
    "    for j in range(len(files)):\n",
    "        y2, sr2 = librosa.load(dirname+\"/\"+files[j])\n",
    "        mfcc2 = librosa.feature.mfcc(y2, sr2)\n",
    "        dist, _, _, _ = dtw(mfcc1.T, mfcc2.T, dist=lambda x, y: norm(x - y, ord=1))\n",
    "#         print files[i],mfcc1.T[0][0],mfcc2.T[0][0],files[j],dist\n",
    "#         if dist<minval:\n",
    "#             minval = dist\n",
    "        distances[i,j] = dist\n",
    "    if i%2==0:\n",
    "        y[i] = 0  #'a'\n",
    "    else:\n",
    "        y[i] = 1  #'b'\n",
    "print(\"Time used: {}s\".format(time.clock()-start))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([   0.        ,  141.22863703,  163.99821788,  192.95202935,\n",
       "        151.85852349,  170.48576354,  135.5725408 ,  152.26686475,\n",
       "        145.76102835,  193.40727232,  137.6427781 ,  161.33156836,\n",
       "        164.65027339,  167.04712037,  176.97619763,  181.20389263,\n",
       "        141.65409762,  168.92626794,  148.07675071,  191.99366183,\n",
       "        152.1781625 ,  193.17680236,  192.36134112,  199.68962323])"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "distances[0] # A dict"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#a = 155.156\n",
    "#b = 184.702\n",
    "#c = 158.231\n",
    "label = ['a','b']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Train a kNN classifier to determine if the audio is 'a' or 'b'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "KNeighborsClassifier(algorithm='auto', leaf_size=30, metric='euclidean',\n",
       "           metric_params=None, n_jobs=1, n_neighbors=3, p=2,\n",
       "           weights='uniform')"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "classifier = KNeighborsClassifier(n_neighbors=3,metric='euclidean')\n",
    "classifier.fit(distances, y)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Test a sample from the test folder"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "y, sr = librosa.load('test/farw0-b1-t.wav')\n",
    "mfcc = librosa.feature.mfcc(y, sr)\n",
    "distanceTest = []\n",
    "for i in range(len(files)):\n",
    "    y1, sr1 = librosa.load(dirname+\"/\"+files[i])\n",
    "    mfcc1 = librosa.feature.mfcc(y1, sr1)\n",
    "    dist, _, _, _ = dtw(mfcc.T, mfcc1.T, dist=lambda x, y: norm(x - y, ord=1))\n",
    "    distanceTest.append(dist)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* Warning: DeprecationWarning: Passing 1d arrays as data is deprecated in 0.17 and will raise ValueError in 0.19. Reshape your data either using X.reshape(-1, 1) if your data has a single feature or X.reshape(1, -1) if it contains a single sample. DeprecationWarning)\n",
    "* [Solution](http://blog.csdn.net/qq_21063873/article/details/60963038)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1.0\n"
     ]
    }
   ],
   "source": [
    "# pre = classifier.predict(distanceTest)[0] # False\n",
    "pre = classifier.predict([distanceTest])[0]\n",
    "print(pre)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'b'"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "label[int(pre)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Predict audio is: 'b'\n"
     ]
    }
   ],
   "source": [
    "print(\"Predict audio is: '{}'\".format(label[int(pre)]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "Python [conda root]",
   "language": "python",
   "name": "conda-root-py"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.5.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}