prepare_cutted_spectrograms.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 169,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import os\n",
    "import pickle\n",
    "import librosa\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import seaborn as sns\n",
    "\n",
    "import scipy.io.wavfile as wav\n",
    "from collections import Counter\n",
    "from matplotlib import pyplot as plt\n",
    "from IPython.display import clear_output\n",
    "\n",
    "from librosa.feature import melspectrogram\n",
    "from librosa import load, logamplitude\n",
    "\n",
    "np.set_printoptions(linewidth=250, suppress=True)\n",
    "\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def find_durations(data, answer):\n",
    "    durations = {}\n",
    "    for i, author in enumerate(set(data.authors)):\n",
    "        for word in data[data.authors == author].words:\n",
    "            #print(os.listdir('dataset/b{}'.format(author)))\n",
    "            wname = 'speech_data/b{}/{}.wav'.format(author, answer[word])\n",
    "            if not os.path.isfile(wname):\n",
    "                fname = 'speech_data/b{}/{}.mp3'.format(author, answer[word])\n",
    "                oname = '/tmp/file.wav'\n",
    "                !lame '{fname}' --decode '{oname}' --quiet\n",
    "                !cp '{oname}' '{wname}'\n",
    "            sample_rate, timesequence = wav.read(wname)\n",
    "            if len(timesequence.shape) > 1:\n",
    "                timesequence = timesequence[:, 0]\n",
    "            durations[(author, word)] = timesequence.size / sample_rate\n",
    "    return durations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "durations = find_durations(data, answer)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 125,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/anaconda3/lib/python3.5/site-packages/statsmodels/nonparametric/kdetools.py:20: VisibleDeprecationWarning: using a non-integer number instead of an integer will result in an error in the future\n",
      "  y = X[:m/2+1] + np.r_[0,X[m/2+1:],0]*1j\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7f26f1d96a20>"
      ]
     },
     "execution_count": 125,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAg8AAAFoCAYAAADQPBjdAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f26f1e26eb8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.distplot(list(durations.values()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Can't init infile 'speech_data/bMshak/2797622731408234712.mp3'\n",
      "Can't init infile 'speech_data/bszurszuncik/-4947795336151861975.mp3'\n",
      "Can't init infile 'speech_data/bszurszuncik/-7183626479791404317.mp3'\n",
      "Can't init infile 'speech_data/bdisappointed/-950460213665267749.mp3'\n",
      "Can't init infile 'speech_data/bdisappointed/6027766216114243336.mp3'\n",
      "Can't init infile 'speech_data/bzhivaya/6142747929449266446.mp3'\n",
      "Can't init infile 'speech_data/b1640max/2324524844955058754.mp3'\n",
      "Can't init infile 'speech_data/b1640max/3063377735457069913.mp3'\n",
      "Can't init infile 'speech_data/bEgde/-3318315914358233874.mp3'\n"
     ]
    }
   ],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 131,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def retrieve_sound_arrays(data, answer, unified_sample_rate=22050):\n",
    "    sequences = {}\n",
    "    #bar = ProgressBar()\n",
    "    for author in set(data.authors):\n",
    "        for word in data[data.authors == author].words:\n",
    "            fname = 'speech_data/b{}/{}.wav'.format(author, answer[word])\n",
    "            #oname = '/tmp/file.wav'\n",
    "            #!lame '{fname}' --decode '{oname}' --quiet\n",
    "            if os.path.isfile(fname):\n",
    "                timesequence, sample_rate = librosa.load(fname)\n",
    "                if len(timesequence.shape) > 1:\n",
    "                    timesequence = timesequence[:, 0]\n",
    "                if sample_rate != 22050:\n",
    "                    unified_timesequence = librosa.resample(np.asarray(timesequence, np.float32),\n",
    "                                                            sample_rate,\n",
    "                                                            unified_sample_rate)\n",
    "                else:\n",
    "                    unified_timesequence = np.array(timesequence, np.float32)\n",
    "                sequences[(author, word)] = timesequence\n",
    "            else:\n",
    "                sequences[(author, word)] = None\n",
    "            \n",
    "    return sequences"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "data = pd.DataFrame.from_csv('data.csv')\n",
    "answer = pickle.load(open('dataset/ajnsdjasnd.pcl','rb'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "durations = find_durations(data, answer)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 118,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3431577249605052561.mp3  3431577249605052561.wav\r\n"
     ]
    }
   ],
   "source": [
    "!ls speech_data/b1999"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 132,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "sequences = retrieve_sound_arrays(data, answer, unified_sample_rate=22050)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 135,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "lengths = np.array([len(arr) for arr in sequences.values() if arr is not None])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "lengths = lengths / 22050"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 138,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "max_duration= 1.0\n",
    "unified_sample_rate=22050\n",
    "n_mels=128\n",
    "flush=True\n",
    "eps=0.0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 139,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "max_samples = int(unified_sample_rate * max_duration)\n",
    "mask = np.ones(len(data), dtype=np.bool)\n",
    "\n",
    "matrix_index_counter = 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 140,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "for index, row in data.iterrows():\n",
    "    next_timesequence = sequences[(row['authors'], row['words'])]\n",
    "    if next_timesequence is None:\n",
    "        # there is no data for that track -- skip it\n",
    "        mask[index] = False\n",
    "    elif len(next_timesequence) / unified_sample_rate < max_duration - eps:\n",
    "        # track is too short -- skip it too\n",
    "        mask[index] = False\n",
    "    else:\n",
    "        # track is ok\n",
    "        continue"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 141,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "valid_data = data.ix[mask]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 146,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "valid_count = len(valid_data)\n",
    "big_matrix = np.zeros((valid_count, max_samples), dtype=np.float32)\n",
    "track_lengths = np.zeros(valid_count, dtype=np.int)\n",
    "matrix_indices = np.arange(valid_count)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 147,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "test_melspectrogram = melspectrogram(np.ones(max_samples), sr=unified_sample_rate, n_mels=n_mels)\n",
    "spectrograms = np.zeros((valid_count,) + test_melspectrogram.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 148,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "track_counter = 0\n",
    "#bar = ProgressBar()\n",
    "for _, row in valid_data.iterrows():\n",
    "    next_timesequence = sequences[(row['authors'], row['words'])]\n",
    "    # hack to handle a little shorter tracks\n",
    "    minsize_clipper = min(next_timesequence.size, max_samples)\n",
    "    big_matrix[track_counter, :minsize_clipper] = next_timesequence[:minsize_clipper]\n",
    "\n",
    "    spectrograms[track_counter, ...] = melspectrogram(np.pad(next_timesequence[:minsize_clipper],\n",
    "                                                     (0, max_samples - minsize_clipper),\n",
    "                                                     mode='constant'),\n",
    "                                              sr=unified_sample_rate, n_mels=n_mels)\n",
    "    track_counter += 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 149,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/anaconda3/lib/python3.5/site-packages/pandas/indexes/api.py:71: RuntimeWarning: unorderable types: str() < int(), sort order is undefined for incomparable objects\n",
      "  result = result.union(other)\n"
     ]
    }
   ],
   "source": [
    "vd = pd.concat([valid_data, pd.Series()])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "pd.Series"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 150,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>words</th>\n",
       "      <th>authors</th>\n",
       "      <th>links</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Бенилюкс</td>\n",
       "      <td>ivga</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%B1%D0%B5%D0%BD%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>большая гармония гласных</td>\n",
       "      <td>rutre15</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%B1%D0%BE%D0%BB%D1...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>малая гармония гласных</td>\n",
       "      <td>rutre15</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%BC%D0%B0%D0%BB%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>бумажный стаканчик</td>\n",
       "      <td>Patient228</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%B1%D1%83%D0%BC%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>бумажный стаканчик</td>\n",
       "      <td>rutre15</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%B1%D1%83%D0%BC%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>будьте любезны</td>\n",
       "      <td>Patient228</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%B1%D1%83%D0%B4%D1...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>будьте любезны</td>\n",
       "      <td>rutre15</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%B1%D1%83%D0%B4%D1...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>улыбчивые</td>\n",
       "      <td>lidadusya</td>\n",
       "      <td>http://ru.forvo.com/word/%D1%83%D0%BB%D1%8B%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>стаканчик мороженого</td>\n",
       "      <td>Patient228</td>\n",
       "      <td>http://ru.forvo.com/word/%D1%81%D1%82%D0%B0%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>стаканчик мороженого</td>\n",
       "      <td>lidadusya</td>\n",
       "      <td>http://ru.forvo.com/word/%D1%81%D1%82%D0%B0%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>экстремизму</td>\n",
       "      <td>lidadusya</td>\n",
       "      <td>http://ru.forvo.com/word/%D1%8D%D0%BA%D1%81%D1...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>гармония гласных</td>\n",
       "      <td>lidadusya</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%B3%D0%B0%D1%80%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>мороженом</td>\n",
       "      <td>szurszuncik</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%BC%D0%BE%D1%80%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>мороженому</td>\n",
       "      <td>szurszuncik</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%BC%D0%BE%D1%80%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>Собор Воскресения Христова на Крови</td>\n",
       "      <td>rutre15</td>\n",
       "      <td>http://ru.forvo.com/word/%D1%81%D0%BE%D0%B1%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>Собор Воскресения Христова на Крови</td>\n",
       "      <td>Patient228</td>\n",
       "      <td>http://ru.forvo.com/word/%D1%81%D0%BE%D0%B1%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>Собор Воскресения Христова на Крови</td>\n",
       "      <td>zibrova_valeriia</td>\n",
       "      <td>http://ru.forvo.com/word/%D1%81%D0%BE%D0%B1%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>арт</td>\n",
       "      <td>Tatar</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%B0%D1%80%D1%82/</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>арт</td>\n",
       "      <td>rutre15</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%B0%D1%80%D1%82/</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>в крови</td>\n",
       "      <td>Patient228</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%B2_%D0%BA%D1%80%D...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>АТС</td>\n",
       "      <td>galari</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%B0%D1%82%D1%81/</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30</th>\n",
       "      <td>аутрайт</td>\n",
       "      <td>Patient228</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%B0%D1%83%D1%82%D1...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31</th>\n",
       "      <td>байга</td>\n",
       "      <td>AqQoyriq</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%B1%D0%B0%D0%B9%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>33</th>\n",
       "      <td>инаугурации</td>\n",
       "      <td>Patient228</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%B8%D0%BD%D0%B0%D1...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34</th>\n",
       "      <td>старшим</td>\n",
       "      <td>Patient228</td>\n",
       "      <td>http://ru.forvo.com/word/%D1%81%D1%82%D0%B0%D1...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35</th>\n",
       "      <td>подписания</td>\n",
       "      <td>Patient228</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%BF%D0%BE%D0%B4%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>37</th>\n",
       "      <td>Азеф</td>\n",
       "      <td>Patient228</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%B0%D0%B7%D0%B5%D1...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>38</th>\n",
       "      <td>автохория</td>\n",
       "      <td>AqQoyriq</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%B0%D0%B2%D1%82%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>39</th>\n",
       "      <td>автохория</td>\n",
       "      <td>Patient228</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%B0%D0%B2%D1%82%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>40</th>\n",
       "      <td>от идеи</td>\n",
       "      <td>Patient228</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%BE%D1%82_%D0%B8%D...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4103</th>\n",
       "      <td>радиочастота</td>\n",
       "      <td>genie_ghoul</td>\n",
       "      <td>http://ru.forvo.com/word/%D1%80%D0%B0%D0%B4%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4104</th>\n",
       "      <td>радиочастотный</td>\n",
       "      <td>suok</td>\n",
       "      <td>http://ru.forvo.com/word/%D1%80%D0%B0%D0%B4%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4105</th>\n",
       "      <td>радиочастотный</td>\n",
       "      <td>genie_ghoul</td>\n",
       "      <td>http://ru.forvo.com/word/%D1%80%D0%B0%D0%B4%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4106</th>\n",
       "      <td>антенно-фидерный</td>\n",
       "      <td>BorisK</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%B0%D0%BD%D1%82%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4107</th>\n",
       "      <td>антенно-фидерный</td>\n",
       "      <td>genie_ghoul</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%B0%D0%BD%D1%82%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4108</th>\n",
       "      <td>волновод</td>\n",
       "      <td>SvetaBerdaus</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%B2%D0%BE%D0%BB%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4109</th>\n",
       "      <td>волновод</td>\n",
       "      <td>genie_ghoul</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%B2%D0%BE%D0%BB%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4111</th>\n",
       "      <td>фидер</td>\n",
       "      <td>genie_ghoul</td>\n",
       "      <td>http://ru.forvo.com/word/%D1%84%D0%B8%D0%B4%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4112</th>\n",
       "      <td>телекоммуникационный</td>\n",
       "      <td>lookyanoff</td>\n",
       "      <td>http://ru.forvo.com/word/%D1%82%D0%B5%D0%BB%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4113</th>\n",
       "      <td>телекоммуникационный</td>\n",
       "      <td>genie_ghoul</td>\n",
       "      <td>http://ru.forvo.com/word/%D1%82%D0%B5%D0%BB%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4115</th>\n",
       "      <td>менеджмент</td>\n",
       "      <td>alexsokol</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%BC%D0%B5%D0%BD%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4116</th>\n",
       "      <td>менеджмент</td>\n",
       "      <td>Eugenia_Ivlieva</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%BC%D0%B5%D0%BD%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4117</th>\n",
       "      <td>менеджмент</td>\n",
       "      <td>1640max</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%BC%D0%B5%D0%BD%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4118</th>\n",
       "      <td>менеджмент</td>\n",
       "      <td>genie_ghoul</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%BC%D0%B5%D0%BD%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4119</th>\n",
       "      <td>менеджмент</td>\n",
       "      <td>AqQoyriq</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%BC%D0%B5%D0%BD%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4120</th>\n",
       "      <td>менеджмент</td>\n",
       "      <td>tarasylo</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%BC%D0%B5%D0%BD%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4121</th>\n",
       "      <td>спектральный</td>\n",
       "      <td>sanych</td>\n",
       "      <td>http://ru.forvo.com/word/%D1%81%D0%BF%D0%B5%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4123</th>\n",
       "      <td>модуляция</td>\n",
       "      <td>alarsso</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%BC%D0%BE%D0%B4%D1...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4126</th>\n",
       "      <td>динамический</td>\n",
       "      <td>sanych</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%B4%D0%B8%D0%BD%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4127</th>\n",
       "      <td>динамический</td>\n",
       "      <td>Maximka</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%B4%D0%B8%D0%BD%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4128</th>\n",
       "      <td>динамический</td>\n",
       "      <td>genie_ghoul</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%B4%D0%B8%D0%BD%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4130</th>\n",
       "      <td>аналоговый</td>\n",
       "      <td>genie_ghoul</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%B0%D0%BD%D0%B0%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4135</th>\n",
       "      <td>волоконно-оптический</td>\n",
       "      <td>Artistee</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%B2%D0%BE%D0%BB%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4136</th>\n",
       "      <td>волоконно-оптический</td>\n",
       "      <td>genie_ghoul</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%B2%D0%BE%D0%BB%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4137</th>\n",
       "      <td>радиосвязь</td>\n",
       "      <td>NeolingvaCom</td>\n",
       "      <td>http://ru.forvo.com/word/%D1%80%D0%B0%D0%B4%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4138</th>\n",
       "      <td>радиосвязь</td>\n",
       "      <td>Maximka</td>\n",
       "      <td>http://ru.forvo.com/word/%D1%80%D0%B0%D0%B4%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4139</th>\n",
       "      <td>радиосвязь</td>\n",
       "      <td>genie_ghoul</td>\n",
       "      <td>http://ru.forvo.com/word/%D1%80%D0%B0%D0%B4%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4140</th>\n",
       "      <td>поливинилхлорид</td>\n",
       "      <td>AqQoyriq</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%BF%D0%BE%D0%BB%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4141</th>\n",
       "      <td>поливинилхлорид</td>\n",
       "      <td>MassEbu</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%BF%D0%BE%D0%BB%D0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4142</th>\n",
       "      <td>поливинилхлорид</td>\n",
       "      <td>genie_ghoul</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%BF%D0%BE%D0%BB%D0...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2677 rows × 3 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                    words           authors  \\\n",
       "0                                Бенилюкс              ivga   \n",
       "1                большая гармония гласных           rutre15   \n",
       "2                  малая гармония гласных           rutre15   \n",
       "3                      бумажный стаканчик        Patient228   \n",
       "4                      бумажный стаканчик           rutre15   \n",
       "5                          будьте любезны        Patient228   \n",
       "6                          будьте любезны           rutre15   \n",
       "7                               улыбчивые         lidadusya   \n",
       "8                    стаканчик мороженого        Patient228   \n",
       "9                    стаканчик мороженого         lidadusya   \n",
       "10                            экстремизму         lidadusya   \n",
       "12                       гармония гласных         lidadusya   \n",
       "13                              мороженом       szurszuncik   \n",
       "14                             мороженому       szurszuncik   \n",
       "16    Собор Воскресения Христова на Крови           rutre15   \n",
       "17    Собор Воскресения Христова на Крови        Patient228   \n",
       "18    Собор Воскресения Христова на Крови  zibrova_valeriia   \n",
       "20                                    арт             Tatar   \n",
       "23                                    арт           rutre15   \n",
       "24                                в крови        Patient228   \n",
       "28                                    АТС            galari   \n",
       "30                                аутрайт        Patient228   \n",
       "31                                  байга          AqQoyriq   \n",
       "33                            инаугурации        Patient228   \n",
       "34                                старшим        Patient228   \n",
       "35                             подписания        Patient228   \n",
       "37                                   Азеф        Patient228   \n",
       "38                              автохория          AqQoyriq   \n",
       "39                              автохория        Patient228   \n",
       "40                                от идеи        Patient228   \n",
       "...                                   ...               ...   \n",
       "4103                         радиочастота       genie_ghoul   \n",
       "4104                       радиочастотный              suok   \n",
       "4105                       радиочастотный       genie_ghoul   \n",
       "4106                     антенно-фидерный            BorisK   \n",
       "4107                     антенно-фидерный       genie_ghoul   \n",
       "4108                             волновод      SvetaBerdaus   \n",
       "4109                             волновод       genie_ghoul   \n",
       "4111                                фидер       genie_ghoul   \n",
       "4112                 телекоммуникационный        lookyanoff   \n",
       "4113                 телекоммуникационный       genie_ghoul   \n",
       "4115                           менеджмент         alexsokol   \n",
       "4116                           менеджмент   Eugenia_Ivlieva   \n",
       "4117                           менеджмент           1640max   \n",
       "4118                           менеджмент       genie_ghoul   \n",
       "4119                           менеджмент          AqQoyriq   \n",
       "4120                           менеджмент          tarasylo   \n",
       "4121                         спектральный            sanych   \n",
       "4123                            модуляция           alarsso   \n",
       "4126                         динамический            sanych   \n",
       "4127                         динамический           Maximka   \n",
       "4128                         динамический       genie_ghoul   \n",
       "4130                           аналоговый       genie_ghoul   \n",
       "4135                 волоконно-оптический          Artistee   \n",
       "4136                 волоконно-оптический       genie_ghoul   \n",
       "4137                           радиосвязь      NeolingvaCom   \n",
       "4138                           радиосвязь           Maximka   \n",
       "4139                           радиосвязь       genie_ghoul   \n",
       "4140                      поливинилхлорид          AqQoyriq   \n",
       "4141                      поливинилхлорид           MassEbu   \n",
       "4142                      поливинилхлорид       genie_ghoul   \n",
       "\n",
       "                                                  links  \n",
       "0     http://ru.forvo.com/word/%D0%B1%D0%B5%D0%BD%D0...  \n",
       "1     http://ru.forvo.com/word/%D0%B1%D0%BE%D0%BB%D1...  \n",
       "2     http://ru.forvo.com/word/%D0%BC%D0%B0%D0%BB%D0...  \n",
       "3     http://ru.forvo.com/word/%D0%B1%D1%83%D0%BC%D0...  \n",
       "4     http://ru.forvo.com/word/%D0%B1%D1%83%D0%BC%D0...  \n",
       "5     http://ru.forvo.com/word/%D0%B1%D1%83%D0%B4%D1...  \n",
       "6     http://ru.forvo.com/word/%D0%B1%D1%83%D0%B4%D1...  \n",
       "7     http://ru.forvo.com/word/%D1%83%D0%BB%D1%8B%D0...  \n",
       "8     http://ru.forvo.com/word/%D1%81%D1%82%D0%B0%D0...  \n",
       "9     http://ru.forvo.com/word/%D1%81%D1%82%D0%B0%D0...  \n",
       "10    http://ru.forvo.com/word/%D1%8D%D0%BA%D1%81%D1...  \n",
       "12    http://ru.forvo.com/word/%D0%B3%D0%B0%D1%80%D0...  \n",
       "13    http://ru.forvo.com/word/%D0%BC%D0%BE%D1%80%D0...  \n",
       "14    http://ru.forvo.com/word/%D0%BC%D0%BE%D1%80%D0...  \n",
       "16    http://ru.forvo.com/word/%D1%81%D0%BE%D0%B1%D0...  \n",
       "17    http://ru.forvo.com/word/%D1%81%D0%BE%D0%B1%D0...  \n",
       "18    http://ru.forvo.com/word/%D1%81%D0%BE%D0%B1%D0...  \n",
       "20         http://ru.forvo.com/word/%D0%B0%D1%80%D1%82/  \n",
       "23         http://ru.forvo.com/word/%D0%B0%D1%80%D1%82/  \n",
       "24    http://ru.forvo.com/word/%D0%B2_%D0%BA%D1%80%D...  \n",
       "28         http://ru.forvo.com/word/%D0%B0%D1%82%D1%81/  \n",
       "30    http://ru.forvo.com/word/%D0%B0%D1%83%D1%82%D1...  \n",
       "31    http://ru.forvo.com/word/%D0%B1%D0%B0%D0%B9%D0...  \n",
       "33    http://ru.forvo.com/word/%D0%B8%D0%BD%D0%B0%D1...  \n",
       "34    http://ru.forvo.com/word/%D1%81%D1%82%D0%B0%D1...  \n",
       "35    http://ru.forvo.com/word/%D0%BF%D0%BE%D0%B4%D0...  \n",
       "37    http://ru.forvo.com/word/%D0%B0%D0%B7%D0%B5%D1...  \n",
       "38    http://ru.forvo.com/word/%D0%B0%D0%B2%D1%82%D0...  \n",
       "39    http://ru.forvo.com/word/%D0%B0%D0%B2%D1%82%D0...  \n",
       "40    http://ru.forvo.com/word/%D0%BE%D1%82_%D0%B8%D...  \n",
       "...                                                 ...  \n",
       "4103  http://ru.forvo.com/word/%D1%80%D0%B0%D0%B4%D0...  \n",
       "4104  http://ru.forvo.com/word/%D1%80%D0%B0%D0%B4%D0...  \n",
       "4105  http://ru.forvo.com/word/%D1%80%D0%B0%D0%B4%D0...  \n",
       "4106  http://ru.forvo.com/word/%D0%B0%D0%BD%D1%82%D0...  \n",
       "4107  http://ru.forvo.com/word/%D0%B0%D0%BD%D1%82%D0...  \n",
       "4108  http://ru.forvo.com/word/%D0%B2%D0%BE%D0%BB%D0...  \n",
       "4109  http://ru.forvo.com/word/%D0%B2%D0%BE%D0%BB%D0...  \n",
       "4111  http://ru.forvo.com/word/%D1%84%D0%B8%D0%B4%D0...  \n",
       "4112  http://ru.forvo.com/word/%D1%82%D0%B5%D0%BB%D0...  \n",
       "4113  http://ru.forvo.com/word/%D1%82%D0%B5%D0%BB%D0...  \n",
       "4115  http://ru.forvo.com/word/%D0%BC%D0%B5%D0%BD%D0...  \n",
       "4116  http://ru.forvo.com/word/%D0%BC%D0%B5%D0%BD%D0...  \n",
       "4117  http://ru.forvo.com/word/%D0%BC%D0%B5%D0%BD%D0...  \n",
       "4118  http://ru.forvo.com/word/%D0%BC%D0%B5%D0%BD%D0...  \n",
       "4119  http://ru.forvo.com/word/%D0%BC%D0%B5%D0%BD%D0...  \n",
       "4120  http://ru.forvo.com/word/%D0%BC%D0%B5%D0%BD%D0...  \n",
       "4121  http://ru.forvo.com/word/%D1%81%D0%BF%D0%B5%D0...  \n",
       "4123  http://ru.forvo.com/word/%D0%BC%D0%BE%D0%B4%D1...  \n",
       "4126  http://ru.forvo.com/word/%D0%B4%D0%B8%D0%BD%D0...  \n",
       "4127  http://ru.forvo.com/word/%D0%B4%D0%B8%D0%BD%D0...  \n",
       "4128  http://ru.forvo.com/word/%D0%B4%D0%B8%D0%BD%D0...  \n",
       "4130  http://ru.forvo.com/word/%D0%B0%D0%BD%D0%B0%D0...  \n",
       "4135  http://ru.forvo.com/word/%D0%B2%D0%BE%D0%BB%D0...  \n",
       "4136  http://ru.forvo.com/word/%D0%B2%D0%BE%D0%BB%D0...  \n",
       "4137  http://ru.forvo.com/word/%D1%80%D0%B0%D0%B4%D0...  \n",
       "4138  http://ru.forvo.com/word/%D1%80%D0%B0%D0%B4%D0...  \n",
       "4139  http://ru.forvo.com/word/%D1%80%D0%B0%D0%B4%D0...  \n",
       "4140  http://ru.forvo.com/word/%D0%BF%D0%BE%D0%BB%D0...  \n",
       "4141  http://ru.forvo.com/word/%D0%BF%D0%BE%D0%BB%D0...  \n",
       "4142  http://ru.forvo.com/word/%D0%BF%D0%BE%D0%BB%D0...  \n",
       "\n",
       "[2677 rows x 3 columns]"
      ]
     },
     "execution_count": 150,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "valid_data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 167,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def build_array_and_dataframe(data, sequences, fname=\"valid_data\", max_duration=0.5,\n",
    "                              unified_sample_rate=22050, n_fft=128, flush=True, eps=0.0):\n",
    "    max_samples = int(unified_sample_rate * max_duration)\n",
    "    mask = np.ones(len(data), dtype=np.bool)\n",
    "        \n",
    "    matrix_index_counter = 0\n",
    "    for index, row in data.iterrows():\n",
    "        next_timesequence = sequences[(row['authors'], row['words'])]\n",
    "        if next_timesequence is None:\n",
    "            # there is no data for that track -- skip it\n",
    "            mask[index] = False\n",
    "        elif len(next_timesequence) / unified_sample_rate < max_duration - eps:\n",
    "            # track is too short -- skip it too\n",
    "            mask[index] = False\n",
    "        else:\n",
    "            # track is ok\n",
    "            continue\n",
    "    \n",
    "    valid_data = data.ix[mask]\n",
    "    valid_count = len(valid_data)\n",
    "    big_matrix = np.zeros((valid_count, max_samples), dtype=np.float32)\n",
    "    track_lengths = np.zeros(valid_count, dtype=np.int)\n",
    "    matrix_indices = np.arange(valid_count)\n",
    "    \n",
    "    # find how many frames return melspectrogram at chosen duration\n",
    "    test_melspectrogram = melspectrogram(np.ones(max_samples), sr=unified_sample_rate, n_fft=n_fft)\n",
    "    spectrograms = np.zeros((valid_count,) + test_melspectrogram.shape)\n",
    "\n",
    "    track_counter = 0\n",
    "    #bar = ProgressBar()\n",
    "    for _, row in valid_data.iterrows():\n",
    "        next_timesequence = sequences[(row['authors'], row['words'])]\n",
    "        # hack to handle a little shorter tracks\n",
    "        minsize_clipper = min(next_timesequence.size, max_samples)\n",
    "        big_matrix[track_counter, :minsize_clipper] = next_timesequence[:minsize_clipper]\n",
    "        \n",
    "        S = melspectrogram(np.pad(next_timesequence[:minsize_clipper],\n",
    "                                                         (0, max_samples - minsize_clipper),\n",
    "                                                         mode='constant'),\n",
    "                                                  sr=unified_sample_rate, n_fft=n_fft)\n",
    "        logS = logamplitude(S, ref_power=np.max)\n",
    "        spectrograms[track_counter, ...] = logS\n",
    "        track_counter += 1\n",
    "        \n",
    "    valid_data.loc[:,'matrix_index'] = pd.Series(matrix_indices, index=valid_data.index)\n",
    "    if flush:\n",
    "        valid_data.to_csv(\"{}.csv\".format(fname))\n",
    "        mat_spec_dict = {\n",
    "            'timesequences': big_matrix,\n",
    "            'spectrograms': spectrograms\n",
    "        }\n",
    "        np.savez('{}'.format(fname), mat_spec_dict)\n",
    "    return valid_data, mat_spec_dict"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 170,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/anaconda3/lib/python3.5/site-packages/pandas/core/indexing.py:296: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy\n",
      "  self.obj[key] = _infer_fill_value(value)\n",
      "/anaconda3/lib/python3.5/site-packages/pandas/core/indexing.py:476: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy\n",
      "  self.obj[item] = s\n"
     ]
    }
   ],
   "source": [
    "vdata, vmatrix = build_array_and_dataframe(data, sequences)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 171,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def retrieve_preprocessed_data(fname='valid_data'):\n",
    "    data_from_file = np.load('{}.npz'.format(fname))\n",
    "    data_table = pd.read_csv('{}.csv'.format(fname))\n",
    "    data_dict = np.asscalar(data_from_file['arr_0'])\n",
    "    return data_table, data_dict['timesequences'], data_dict['spectrograms']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 172,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "valid_data, valid_timesequences, valid_spectrograms = retrieve_preprocessed_data('valid_data')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 160,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 1.05904762,  1.32027211,  1.71210884, ...,  1.73823129,  1.60761905,  0.56272109])"
      ]
     },
     "execution_count": 160,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lengths"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 161,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(4088, 128, 22)"
      ]
     },
     "execution_count": 161,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "valid_spectrograms.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 159,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'vdata_from_file' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-159-858fe99fc826>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mtype\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0masscalar\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mvdata_from_file\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'arr_0'\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m: name 'vdata_from_file' is not defined"
     ]
    }
   ],
   "source": [
    "type(np.asscalar(vdata_from_file['arr_0']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 144,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>words</th>\n",
       "      <th>authors</th>\n",
       "      <th>links</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>6767</th>\n",
       "      <td>посещать</td>\n",
       "      <td>arina</td>\n",
       "      <td>http://ru.forvo.com/word/%D0%BF%D0%BE%D1%81%D0...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         words authors                                              links\n",
       "6767  посещать   arina  http://ru.forvo.com/word/%D0%BF%D0%BE%D1%81%D0..."
      ]
     },
     "execution_count": 144,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.loc[np.array((data['authors'] == lst[0][0]) & (data['words'] == lst[0][1]))]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 130,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "ename": "KeyError",
     "evalue": "0",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mKeyError\u001b[0m                                  Traceback (most recent call last)",
      "\u001b[1;32m/root/anaconda3/lib/python3.5/site-packages/pandas/indexes/base.py\u001b[0m in \u001b[0;36mget_loc\u001b[1;34m(self, key, method, tolerance)\u001b[0m\n\u001b[0;32m   1875\u001b[0m             \u001b[1;32mtry\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 1876\u001b[1;33m                 \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_engine\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mget_loc\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   1877\u001b[0m             \u001b[1;32mexcept\u001b[0m \u001b[0mKeyError\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mpandas/index.pyx\u001b[0m in \u001b[0;36mpandas.index.IndexEngine.get_loc (pandas/index.c:4027)\u001b[1;34m()\u001b[0m\n",
      "\u001b[1;32mpandas/index.pyx\u001b[0m in \u001b[0;36mpandas.index.IndexEngine.get_loc (pandas/index.c:3891)\u001b[1;34m()\u001b[0m\n",
      "\u001b[1;32mpandas/hashtable.pyx\u001b[0m in \u001b[0;36mpandas.hashtable.PyObjectHashTable.get_item (pandas/hashtable.c:12408)\u001b[1;34m()\u001b[0m\n",
      "\u001b[1;32mpandas/hashtable.pyx\u001b[0m in \u001b[0;36mpandas.hashtable.PyObjectHashTable.get_item (pandas/hashtable.c:12359)\u001b[1;34m()\u001b[0m\n",
      "\u001b[1;31mKeyError\u001b[0m: 0",
      "\nDuring handling of the above exception, another exception occurred:\n",
      "\u001b[1;31mKeyError\u001b[0m                                  Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-130-a8c72f481711>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m()\u001b[0m\n\u001b[1;32m----> 1\u001b[1;33m \u001b[0mdata\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;36m1\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;36m0\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[1;32m/root/anaconda3/lib/python3.5/site-packages/pandas/core/frame.py\u001b[0m in \u001b[0;36m__getitem__\u001b[1;34m(self, key)\u001b[0m\n\u001b[0;32m   1990\u001b[0m             \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_getitem_multilevel\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1991\u001b[0m         \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 1992\u001b[1;33m             \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_getitem_column\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   1993\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1994\u001b[0m     \u001b[1;32mdef\u001b[0m \u001b[0m_getitem_column\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mkey\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m/root/anaconda3/lib/python3.5/site-packages/pandas/core/frame.py\u001b[0m in \u001b[0;36m_getitem_column\u001b[1;34m(self, key)\u001b[0m\n\u001b[0;32m   1997\u001b[0m         \u001b[1;31m# get column\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1998\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mcolumns\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mis_unique\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 1999\u001b[1;33m             \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_get_item_cache\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   2000\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   2001\u001b[0m         \u001b[1;31m# duplicate columns & possible reduce dimensionality\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m/root/anaconda3/lib/python3.5/site-packages/pandas/core/generic.py\u001b[0m in \u001b[0;36m_get_item_cache\u001b[1;34m(self, item)\u001b[0m\n\u001b[0;32m   1343\u001b[0m         \u001b[0mres\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mcache\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mget\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mitem\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1344\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[0mres\u001b[0m \u001b[1;32mis\u001b[0m \u001b[1;32mNone\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 1345\u001b[1;33m             \u001b[0mvalues\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_data\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mget\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mitem\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   1346\u001b[0m             \u001b[0mres\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_box_item_values\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mitem\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mvalues\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1347\u001b[0m             \u001b[0mcache\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mitem\u001b[0m\u001b[1;33m]\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mres\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m/root/anaconda3/lib/python3.5/site-packages/pandas/core/internals.py\u001b[0m in \u001b[0;36mget\u001b[1;34m(self, item, fastpath)\u001b[0m\n\u001b[0;32m   3223\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   3224\u001b[0m             \u001b[1;32mif\u001b[0m \u001b[1;32mnot\u001b[0m \u001b[0misnull\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mitem\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 3225\u001b[1;33m                 \u001b[0mloc\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mitems\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mget_loc\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mitem\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   3226\u001b[0m             \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   3227\u001b[0m                 \u001b[0mindexer\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mnp\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0marange\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mlen\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mitems\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0misnull\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mitems\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m/root/anaconda3/lib/python3.5/site-packages/pandas/indexes/base.py\u001b[0m in \u001b[0;36mget_loc\u001b[1;34m(self, key, method, tolerance)\u001b[0m\n\u001b[0;32m   1876\u001b[0m                 \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_engine\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mget_loc\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1877\u001b[0m             \u001b[1;32mexcept\u001b[0m \u001b[0mKeyError\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 1878\u001b[1;33m                 \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_engine\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mget_loc\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_maybe_cast_indexer\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   1879\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1880\u001b[0m         \u001b[0mindexer\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mget_indexer\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mmethod\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mmethod\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mtolerance\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mtolerance\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mpandas/index.pyx\u001b[0m in \u001b[0;36mpandas.index.IndexEngine.get_loc (pandas/index.c:4027)\u001b[1;34m()\u001b[0m\n",
      "\u001b[1;32mpandas/index.pyx\u001b[0m in \u001b[0;36mpandas.index.IndexEngine.get_loc (pandas/index.c:3891)\u001b[1;34m()\u001b[0m\n",
      "\u001b[1;32mpandas/hashtable.pyx\u001b[0m in \u001b[0;36mpandas.hashtable.PyObjectHashTable.get_item (pandas/hashtable.c:12408)\u001b[1;34m()\u001b[0m\n",
      "\u001b[1;32mpandas/hashtable.pyx\u001b[0m in \u001b[0;36mpandas.hashtable.PyObjectHashTable.get_item (pandas/hashtable.c:12359)\u001b[1;34m()\u001b[0m\n",
      "\u001b[1;31mKeyError\u001b[0m: 0"
     ]
    }
   ],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "durations = find_durations(data, answer)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 162,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "dur_values = np.array(list(durations.values()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 163,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/anaconda3/lib/python3.5/site-packages/statsmodels/nonparametric/kdetools.py:20: VisibleDeprecationWarning: using a non-integer number instead of an integer will result in an error in the future\n",
      "  y = X[:m/2+1] + np.r_[0,X[m/2+1:],0]*1j\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7f26e7abf978>"
      ]
     },
     "execution_count": 163,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA0UAAAInCAYAAABXzPJQAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f26e7abfdd8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.plt.figure(figsize=(10, 6))\n",
    "sns.plt.title('distribution of sound lengths')\n",
    "sns.plt.xlabel('time, sec')\n",
    "sns.distplot(lengths)\n",
    "#sns.plt.savefig('soundlength_distribution.png', dpi=150)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 164,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6149659863945578"
      ]
     },
     "execution_count": 164,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.percentile(lengths, 5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "3.2271882086167802"
      ]
     },
     "execution_count": 72,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dur_values.max()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "data = pd.DataFrame.from_csv('data.csv')\n",
    "answer = pickle.load(open('dataset/ajnsdjasnd.pcl','rb'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "ex_author = data.authors[0]\n",
    "ex_word = data[data.authors == ex_author].words[0]\n",
    "fname = 'dataset/b{}/{}.mp3'.format(ex_author, answer[ex_word])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "ename": "NoBackendError",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mNoBackendError\u001b[0m                            Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-14-2bd4d882a1c5>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m()\u001b[0m\n\u001b[1;32m----> 1\u001b[1;33m \u001b[0mlibrosa\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mload\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mfname\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[1;32m/root/anaconda3/lib/python3.5/site-packages/librosa/core/audio.py\u001b[0m in \u001b[0;36mload\u001b[1;34m(path, sr, mono, offset, duration, dtype)\u001b[0m\n\u001b[0;32m    107\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    108\u001b[0m     \u001b[0my\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;33m[\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 109\u001b[1;33m     \u001b[1;32mwith\u001b[0m \u001b[0maudioread\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0maudio_open\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mos\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mpath\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mrealpath\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mpath\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m \u001b[1;32mas\u001b[0m \u001b[0minput_file\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    110\u001b[0m         \u001b[0msr_native\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0minput_file\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0msamplerate\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    111\u001b[0m         \u001b[0mn_channels\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0minput_file\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mchannels\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m/root/anaconda3/lib/python3.5/site-packages/audioread/__init__.py\u001b[0m in \u001b[0;36maudio_open\u001b[1;34m(path)\u001b[0m\n\u001b[0;32m    112\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    113\u001b[0m     \u001b[1;31m# All backends failed!\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 114\u001b[1;33m     \u001b[1;32mraise\u001b[0m \u001b[0mNoBackendError\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[1;31mNoBackendError\u001b[0m: "
     ]
    }
   ],
   "source": [
    "librosa.load(fname)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def get_spectgorgamm(fname): \n",
    "    y, sr = librosa.load(fname) \n",
    "    S = librosa.feature.melspectrogram(y, sr=sr, n_mels=128) \n",
    "    log_S = librosa.logamplitude(S, ref_power=np.max) \n",
    "    return log_S[:,:1200]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 159,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2149.1146087646484"
      ]
     },
     "execution_count": 159,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "3.5 * 22050 * 4 * 7300 / 1024 / 1024"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'predict' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-2-19be86c8c029>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mpredict\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m: name 'predict' is not defined"
     ]
    }
   ],
   "source": [
    "predict()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python [default]",
   "language": "python",
   "name": "python3"
  },
  "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": 0
}