LiveDemo.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# KeyWordSpotting - LIVE DEMO"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Imports\n",
    "import librosa\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import sounddevice as sd\n",
    "import Models #Our models\n",
    "import LoadAndPreprocessDataset\n",
    "\n",
    "categories=['yes','no','up','down','left','right','on','off','stop','go']\n",
    "#categories=['yes','no','up','down','left','right','on','off','stop','go','zero','one','two','three','four','five','six','seven','eight','nine','unknown']\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1) Audio Input\n",
    "### Recording from microphone"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "sr = 16000\n",
    "seconds = 10\n",
    "\n",
    "myrec = sd.rec(int(sr*seconds), samplerate=sr, channels=1)\n",
    "sd.wait()\n",
    "\n",
    "np.save('LiveDemo/myrec2.npy', myrec)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Reading it from a numpy array"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "sr=16000\n",
    "#myrec = np.load(\"LiveDemo/ONaudioclip.npy\")\n",
    "myrec = np.load(\"LiveDemo/myrec.npy\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Reading it from a wav file"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "myrec, sr = librosa.load(\"LiveDemo/audioclip.wav\", sr=None)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Audio Clip Informations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Sample rate =\t16000 samples/s\n",
      "Duration t =\t10.0 s\n",
      "Average intensity = 0.009745188\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA5MAAAFzCAYAAACjNF0iAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1 /AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy WH4yJAAAgAElEQVR4nOzdeXxU1f3/8fcne0jCmrCDYUcEQcVdURTc69LaVmtbbW3Vtlqr1m 129cuKl1 rX5braXWpdbW3WoVN3BBBdkUQZB9XxMIZN9zfn9kEkNIQiaZmTMzeT0fjzyYuXNz7zvcyeR 7jn3HHPOCQAAAACAYCT4DgAAAAAAiD0UkwAAAACAoFFMAgAAAACCRjEJAAAAAAgaxSQAAAAAIGgUkwAAAACAoCX5DhAO2dnZLjc313cMAAAAAPBiyZIle5xzOeHcR1wWk7m5uVq8eLHvGAAAAADghZltDvc 6OYKAAAAAAgaxSQAAAAAIGgUkwAAAACAoFFMAgAAAACCRjEJAAAAAAgaxSQAAAAAIGgUkwAAAACAoFFMAgAAAACCRjEJAAAAAAgaxSQAAAAAIGhei0kzO8fMVpvZOjO7rY31jjWzWjO7NJL5AAAAAAAt81ZMmlmipPsknStpnKTLzWxcK v9RtJrkU0IAED82binVDW1db5jAADigM WyeMkrXPObXDOVUl6QtJFLax3g6RnJeVFMhwAAPFmZ2G5pv7 bd0561PfUQAAccBnMTlI0tYmz7cFljUys0GSLpH0wKE2ZmbXmNliM1ucn58f0qAAAMS6eev26MS735QkPfz Jr9hAABxwWcxaS0sc82e3yPpR8652kNtzDk30zk32Tk3OScnJyQBAQCIF6 v3O07AgAgziR53Pc2SUOaPB8saUezdSZLesLMJClb0nlmVuOc 09kIgIAAAAAWuKzmFwkaZSZDZO0XdJlkr7SdAXn3LCGx2b2iKSXKCQBAAietdQfCACATvBWTDrnaszsetWP0poo6SHn3Aozuy7w iHvkwQAAO1jLd5dAgBAx/lsmZRzbpakWc2WtVhEOueuikQmAAAAAMCh RyABwAARAjdXAEAoUYxCQBAF0AtCQAINYpJAAAAAEDQKCYBAOgC6OYKAAg1ikkAALoAo5oEAIQYxSQAAAAAIGgUkwAAdAG0SwIAQo1iEgCAroBqEgAQYhSTAAAAAICgUUwCANAFGE2TAIAQo5gEAKALYDBXAECoUUwCAAAAAIJGMQkAQBdAwyQAINQoJgEA6ALo5goACDWKSQAAAABA0CgmAQDoAlbvKvYdAQAQZygmAQDoAmZ/muc7AgAgzlBMAgAAAACCRjEJAAAAAAgaxSQAAAAAIGgUkwAAAACAoFFMAgAAAACCRjEJAAAAAAgaxSQAAAAAIGgUkwAAAACAoFFMAgAQ52rrnO8IAIA4RDEJAECcu2f2moOWLdu230MSAEA8oZgEACDOLd16cOF44Z/f15LN zykAQDEC4pJAAC6qJ2F5b4jAABiGMUkAABdVFVNne8IAIAYRjEJAEAXdfNTH/uOAACIYRSTAAAAAICgUUwCAAAAAILmtZg0s3PMbLWZrTOz21p4/SIzW2ZmS81ssZmd4iMnAAAAAOBA3opJM0uUdJ kcyWNk3S5mY1rttocSROdc5MkfVPSg5FNCQBAfPvRM8t8RwAAxCifLZPHSVrnnNvgnKuS9ISki5qu4Jwrcc65wNMMSU4AACBknly81XcEAECM8llMDpLU9C/YtsCyA5jZJWa2StLLqm dbJGZXRPoCrs4Pz8/5GEBAAAAAJ/xWUxaC8sOanl0zj3vnBsr6WJJv2ptY865mc65yc65yTk5OSGMCQAAAABozmcxuU3SkCbPB0va0drKzrm5kkaYWXa4gwEA0JXkFVX4jgAAiEE i8lFkkaZ2TAzS5F0maQXm65gZiPNzAKPj5aUImlvxJMCABDHjrtrju8IAIAYlORrx865GjO7XtJrkhIlPeScW2Fm1wVef0DSFyR93cyqJZVL nKTAXkAAEA7BK7LtmnVriKN7d89AmkAAPHCWzEpSc65WZJmNVv2QJPHv5H0m0jnAgCgq9myt4xiEgAQFJ/dXAEAQJSg2w8AIFgUkwAAAACAoFFMAgAAMSIBACBYFJMAAAAAgKBRTAIAAHHXJHxal1es6to63zEABIliEgCAOLe/rMp3BKBVWwvKNO0Pc3X3rFW owAIEsUkAABxbtm2wkOuwz2T8KWgtP5ix LNBZ6TAAgWxSQAAKCTKwAgaBSTAAAAAICgUUwCAAB99/EPtWzbft8xAAAxhGISAABIku6dvdZ3BABADKGYBAAAAAAEjWISAAAAABA0ikkAAOJYw7QLAACEGsUkAABx6vUVu3T0r97wHQMAEKcoJgEAiFMLNzIJPAAgfCgmAQAAAABBo5gEAAAAAASNYhIAAAAAEDSKSQAAAHi3bFuh7wgAgkQxCQAAAAAIGsUkAAAAACBoFJMAAAAAgKBRTAIAAAAAgkYxCQBAnDLznQAAEM8oJgEAAAAAQaOYBAAAAAAEjWISAAAAABA0ikkAACBJcr4DIO5V1tTqx88vV0Fple8oAEKAYhIAAAAR8eLSHfrXgi26e9anvqMACAGKSQAAAITdmt3FuvWZZZKkoopqz2kAhALFJAAAAMLum48sanz81up8j0kAhIrXYtLMzjGz1Wa2zsxua H1K8xsWeBrnplN9JETAAAA4Xfzk0t9RwAQBG/FpJklSrpP0rmSxkm63MzGNVtto6TTnHNHSvqVpJmRTQkAAIBIee6j7b4jAAiCz5bJ4yStc85tcM5VSXpC0kVNV3DOzXPO7Qs8/UDS4AhnBIBOc87pwy37Dr0iAHQRVTV1WrypQJJUXVvnOQ2AjvJZTA6StLXJ822BZa25WtIrYU0EAGHw5KKt vz98/TqJ7t8R0EXsy6vxHcEoFWXPjBfFdW1uvSB b6jAOigJI/7thaWtTjFlZlNVX0xeUqrGzO7RtI1kjR06NBQ5AOAkGg4od9aUOY5CboaBjlBtKtzzG4KxDKfLZPbJA1p8nywpB3NVzKzIyU9KOki59ze1jbmnJvpnJvsnJuck5MT8rAAAAAIrSWbuQUAiGU i8lFkkaZ2TAzS5F0maQXm65gZkMlPSfpa865NR4yAgAAIARaaoT82t8XRj4IgJDx1s3VOVdjZtdLek1SoqSHnHMrzOy6wOsPSPq5pD6S7jczSapxzk32lRkAgHjW0v0nAAC0xuc9k3LOzZI0q9myB5o8/pakb0U6FwCEg2v5tnAAAICY5LObKwB0CRZo7rlr1iq/QQAAAEKIYhIAImhXYYXvCAAAACFBMQkAEXTC3XNUW0d3VwBdj3FTLhB3KCYBIMLqnFNlTa0c86shysxZlec7AgAghlBMAkCY7S6qPOD5rsIKjfnpq3rsg82eEgFA5HH9DIg/FJMAEGYvfrzjgOfr8kskST9/YYWPOECbCkqrtLOw3HcMxLCv/X2Bfv3SSt8xAEQAxSQARNg3Hl7kOwLQqqN/9YZOvPtN3zEQw95du0cPvrfRdwwAEUAxCQAAgJBoOsDY39/bqN1FjGANxDOKSQAA4tD2/XRVRWTV1TmN PGsxue/emmlrnlsicdEAMKNYhIAgDh08gy6qiKySqpqDlr28db9kqSPtuzjAgcQhygmAQAA0GkPvL2 xeVb9pbpkvvntX8776xn6iQgRlBMAgAAoNPub6WYnPK7t4LazoxXVunDLftDEQlAmFFMAgAAIKo0HcgHQPSimASAMNpaUNbm65wwIVpt3FPqOwIAIMpRTAJAGH3pr/PbfP2afyyOUBIgOGUtDKYCAEBTFJMAEEb7y6rbfH3OqrwIJQGCYzLfERBD5q3fE9LtMQAPEBsoJgEAwEH 9u4G5RUz4TzaZ/m2Qt8RAHhAMQkAAA7y/EfbdcO/PvIdAwAQxSgmAQBAi4oruG8SfqzPL1V1bZ3vGAAOgWISAMLIuO0MMWzlziLfERAjQv1Z9 Pnl vuWatCu1EAIUcxCQAAWrWnpNJ3BMSA5z7cHvJtLt5cEPJtAggtikkAANCq8qpa3xEQ5Z5avFWrdhX7jgHAA4pJAADQqiruW8Mh/PnNdb4jAPCEYhIAALTqK3/7wHcEAECUopgEAACt2l3EPZPwwznfCQAcCsUkAAA4pA35JXps/ibfMRBlnHPaUlDmOwYAT5J8BwCAeFZZw/1miLy3VuWFfJuX3D9PheXVuuL4w5SQwJw3kOrqnH772mrfMQB4RMskAIRRbR39tBB5/1ka mkaCsurQ75NxLbHF27RA  s9x0DgEcUk4g5WwvKVFjGSQ0ARMofXqf1CQfbti 83VuduBgHRDuKScSUmto6nfrbt3T2PXN9RwEOqbiifRc9lm7dH Yk6GpCPXDJE4u2Nj7eUlCm3Nte1u3PLQvtTgAAMYdiEjHl5y ukCTtKqrwnAQ4tBc/3tGu9f75weYwJwE6p2lt u9FW r/Xbi15ZXRZZi4dxbo6igmEVNeanJyvmpXkcckwKG1916iZ5ZsC3MSdDUfbNgbtm3/9Z0NjY//NndDG2siXr22YpeeXszFBAAUk4hhq3cV 44AtGlrQbnvCOii8opDOzdkfivbu3PWpyHdD2LDtY8t0a3PLFMdE0ECXZ7XYtLMzjGz1Wa2zsxua H1sWY238wqzeyHPjICAADgYDPD3DLdUKu sHS7/rVgS1j3BaBjvBWTZpYo6T5J50oaJ lyMxvXbLUCSd X9PsIxwOAiJq3bo/vCAAQlW58Yql /Pxy3zEAtMBny RxktY55zY456okPSHpoqYrOOfynHOLJDEPBA7yhzfW I4AhMxXHlzgOwIARBVjfB8g6vksJgdJanr39rbAsg4xs2vMbLGZLc7Pz 90OEQna/KXZfPeMqZUAADP3lqV5zsC4lRldZ3vCAAOwWcx2dL1pg7fye2cm mcm ycm5yTk9OJWIglLy5t39QLQKRt2lMa9Pesy2NQKcSeJxcxqifCY21eie8IAA7BZzG5TdKQJs8HS6IyABAX/vZu8ANT/PMDBpgAgKb2lVb5jgCgDe0qJs2sdxj2vUjSKDMbZmYpki6T9GIY9oM4cPesT/XNRxapsJzbZxG/3lpNd0EAaKqyhq6uQDRrb8vkAjN72szOMwvN7dDOuRpJ10t6TdKnkp5yzq0ws vM7DpJMrP ZrZN0s2Sfmpm28yseyj2j9jy17kb9GYL9 U89P5GD2mA8Ni8t8x3BCBor67Y1eLyyppaFZZxARCdc9KMOb4jAGhDUjvXGy1pmqRvSvqTmT0p6RHnXKeG03TOzZI0q9myB5o83qX67q8AELXq6pzmrs3XaaNzFKLrbUBMcc4d9N7/2t8XauHGAm2acb6nVIgHdR0eTQNAJLSrZdLVe8M5d7mkb0m6UtJCM3vHzE4Ma0IAiEIPvbdRs1fuliQ9On Trnp4kV5evlOSVFFdq8c7OMH2zsLyUEUEImbY7bNU1 ysf HGAk9pEE58RgFoqr33TPYxsxvNbLGkH0q6QVK2pFsk/SuM QCtz2c0N0SfX760Ut/6x2JJ0taC pOrXYUVkqRrHlvS4e1 8YH5nQ8HePDA3PW IyACTrz7Td8RAESR9nZznS/pMUkXO e2NVm 2MweaOV7gJBo79Xtmto6lVTWqGe3lDAnAg7UvGfr3DUdn t2 36u iM2Ld3S8ry/BaVV6p3B5zIAxKP2DsDzU fcr5oWkmb2RUlyzv0mLMmAdtpdVN8a9PMXV2jSL99QRXWt50ToSpZsLlBRYJRhF4J7e5yTthaU6Xv/ lCPMMAU4sAJdzOACgDEq/YWk7e1sOz2UAYBOur4u aosLxa/11aP01pVS3DiCNy9pZU6ekl2w69YhDW5Zfo5WU7dcd/V4Z0u0AklFTWaNWuosbnVTV1jOoKAHGqzW6uZnaupPMkDTKz/2vyUndJNeEMBgRjXV6Jiit5SyLymt4fmV9SqXtmd2qQa0lS016zd7/yqW47ZyyjxCLqLdpUoIUbC/Slvx583 85987V/NvPbHy u6hC2ZmpSkzgfQ0AsexQ90zukLRY0oWSmo4oUSzppnCFAoL1hb/Ma3wciq6GQFteXrazxeUz524Iyfb3lVU1Pv7rOxv0tRMO0 Be3UKybSBc9pVVt1hIStLOwgpt2lOqf36wWY/M26SaOqerTxmmn10wLsIpAQCh1GYx6Zz7WNLHZva4c45mHwCQ9MSijk370V43PfnxAc9fX7Fb3zxlWFj3CYTb6b9/ 4Dnb6zcTTGJoBSWV6tHerLvGACaaPOeSTN7KvDwIzNb1uRruZkti0A QE8v3uo7AnAAupwCnVfbZF7Ku2Z9qiWb93lMg1hw5UMLfUcA0MyhurneGPj3gnAHAVqyZnexPmxluHmgq6ipY1AptKy6tk7OSSlJ7R1PL3oUlVeruKJaWWnJmjl3g2bO3aBNM873HQtRbOlWzgeAaNPmXx/nXMONQXskbXXObZaUKmmi6u nBMLq2Q9DO0omEAqRHjPkrlmrtHpXcWR3iqh381NLNeonr2j0T1/xHaVDiitrNOGO133HAAB0QnsvZc6VlGZmgyTNkfQNSY EKxTQKQzAgzDz0cn1mscWe9grotlzH273HQFdiHNOpYyaDqCZQ3VzbWDOuTIzu1rSn5xzvzWzj8IZDJDqR7IEIG3eW6Z9pVXqlZHiOwoQUs/RAyUmXPrAfO5rBXCQ9rZMmpmdKOkKSS8HlrW3EAUiqriSybERXr4G4NldXOFlv4g 76zJ9x0hZP73xRW I6AdKCQBtKS9xeSNkm6X9LxzboWZDZf0VvhiAR335b9 4DsCEBZvrYqfAgKd89j8za2 9sv/roxgks6rrq0fYIpBkgEg9rSrmHTOzXXOXeic 03g Qbn3PfDGw3omO37y UcN04ifHyd8/7m1VWe9oxY8tD7G31HCErDDCHUkgAQe9pVTJrZaDObaWavm9mbDV/hDgd01OMLwjupPLo2ny0o76/b42/niFofbtmnlTuKfMfokIaLf3VOKq7gNgUAiCXtve/xaUkPSHpQUm344gCfqavreOvi1n1lIUwCRI8rHlzAXHxQ82GrP3//PEnSu/8z1UeYTqmu/exnWbO7RMcc1stjGkS7iupapSUn o4BIKC9xWSNc 4vYU0CNPPA3PUd/l6jwxTCivcXotPH25jUHfHtzVV5Om/CAN8xAAS0dwCe/5rZd81sgJn1bvgKazJ0eb99dXWHv7e8irmwAMSz LygwSA8OJSSCMx1uXFPqbbvLw/7foB40N5i8kpJt0qaJ2lJ4IsZtBG1Hm1jpEMAiH0t3wZw/b9iewro0ggUCohtDaP/htPU37 tk2cwNAjQHu0dzXVYC1/Dwx0OAAB0HV/7 0LfERDlfvL8J74jAGiivaO5djOzn5rZzMDzUWZ2QXijoSuLxJVHoKPoigcAAND bq4PS6qSdFLg TZJvw5LohCorXNau7vYdwx0wh0vrvAdAWiV71oyr7jCcwIAAID2F5MjnHO/lVQtSc65cvk/n2rV vwSTf/jXN8x0AFLNu9TRXVtSOaJrKiu1aJNBSFIBRxo1S6/F6se457gLs91fOakqFdTW6fSyhpV1dBDBQCiXXunBqkys3QF7vg3sxGSKsOWqpMq QMUc7bvLw/5ze5jf/aqpPp514b07hbSbaNr21Lgdx7TP725TrecNcZrBiBcRv7kFUnSsbm99PR1Jx1ibURCZ Z9Dod31 br1FE5vmMAUPtbJu Q9KqkIWb2uKQ5kn4UrlDoema 0/E5JQ luILRARF/XDw3TeGQarvA8V 0aZ/vCAi4Z85a3xEOwEBNQPRoV8ukc 51M1si6QTVd2 90Tm3J6zJ0KWEcyoPBktBPKqorlN6SqLvGPDk7dX5viOgC8kvjtrOaAA8a 9ornOcc3udcy87515yzu0xsznhDgeEwrn3vus7AhByr63Y5TsCEHb3zl6r99dx7dq3zNSue Gqrs5pzqe76Q0CtKLNYtLM0syst6RsM tlZr0DX7mSBkYiIOLfPE4UgKD94Mmlqqyp9R0DCKs/zl6jKx5c4DtGlxdlt0xKksqqInMLyzNLtunqRxfrqcVbI7I/INYcqmXyWklLJI0N/Nvw9YKk 8IbDV3Fks3hvy8m97aXtXJHUdj3A0QSU igq1i6dT9Tfnn03IfbfEc4yK1PLwv7PpZsLtDavPr33WMfMIp2c6WVNcorYqqqrq7NYtI5d69zbpikHzrnhjvnhgW Jjrn/hyhjIhzCQmRuanxvP97V7m3vRzU9zjn5Fz9vKXb95eHKRliyRsrd/uO0OjdtfWt mVVXecPem2do7tZF3Txfe9HfMqvhilK2lJQWtU40ml1bZ0qqjvfW2B/WVWntxFKFdW12ldW7TvGQV5evlN5RRX6/Wur9c6afG3cU9rm h9t2acVOwqD2scX/jJff3t3oyTpk 1tX5AurazRH99Yo rajs0oUFFd2 7PtlnLd qV5TsPWl5YXq0lm/epsLxa763do8cXbFbubS 3 J768fPLtXZ3sZ5dsk0vL9upy2bOV2F5tf778Y5W91tdW6eqmjrtL6vSzsJynfXHuTrurvbf9bZxT6nW7i7W26vz9MLS7Ydc/8lFWzSryc85d02 XmwjX4N9pVUqaeN3t6yqRs8s2db4//2T55cr97aXlVdUoVeW79Sa3cUh/T1csrlA89fvlVT/uXLP7DUqrjj079T sirlFVeovKp9nyufbC/Ux1v3S6o/f90RofPW9g7A8yczO0lSbtPvcc79ozM7N7NzJN0rKVHSg865Gc1et8Dr50kqk3SVc 7DzuwT0aegNLJ/OD/ZXqgL/vTeQcsTTEpOTGhzapn7vnK0po3rq6SEBCUmmD7ZXqjR/bKUktTegZFjV1FFtZISTN1SWv7Y2Ly3VEN7d5OZKa oQvvLq1VVU6fVu4o1oGeaThqRHdZ8m/eW6ux75uqjn5110MA0zjmZmWpq65SYYLI2RmXaWlCm5z7crmumDG9xgJv/feGTkGfvqG37yg 4QDIsO0N/ erRWrO7RN3TkjRlVI4SEkx1dU7/Wbpdn5s4UMmJB79X6y afHZhZ966PZowuIcSWznem/aUqqSyRgN7pmvbvjId1idDPdKTJUl7SypVVVunAT3Sm QsU89uKcpM/Wxbc9fkq6yqfi7Y4TkZ2ldapa ecJheWrZTF04aqEUbC3Tm4f2UV1Sh Rv26rwJA7R9X7lO//3b sWFR jKk3IlSSWVNTLVn6SMH9RDkvT8R9t08ohs9e2e1ri/wrJqZaQmKikxQbuLKtSvyWsn3T1HN581RiP7ZiopwVRb5zRxSM/G13/x3xUa2rubvnL8UDlXPzXMyJxMrc8vUZ/MVGWlJWlPSaUG9EiXc07r80uUmpSo1KQEJSUmqHdGimpq65QfWKe0ska7iio0IiezcR9VNXVyckpNavm9m1dUoe7pyUpLTtQHG/a2/qaIc//zzMe65KjB njbfp03foD uWCzrjwpV8kJpuc 2q5rpww/6Pe7tLJG3VISG5c75/Thln0a3S9La/NKNKBHmhITTBkpSUpPTmz8Pfj2PxbrrdX52jTjfK3LK9Gekkr9e EWvbB0h 66ZILeXLVbsz/N0w1njNTa3SWa/elu1dQ5zbvtDA3okaaFGwt01NBeSk40FVfWKCUxQUkJpqTEBG0tKFOPbsnqnlb/e1NcUa11eSUqLK/WVQ8v0p8uP0qfm9jy3USrdxUrNSlBPdKTVV1bpxU7izR1TF9J0oINezWoV7oG9 qmmto6LdxUoP7d09S3e5rSkur/bjknLd22X7l9MtQ7I6VxTuaGqTb2l1Vpze76/5cBPdL00/9Ez2dec42FzFsHLr/93LGaOXeDjjmslx746jGav2FvY3fpB756jM4 ol/j  E3r67SF44epMSEBPXJTGk8GW9J7m0va9OM87ViR6HGDeiulTuL1CM9WXtLqnTNY4u1u6hS1bV1uv/t9Tp1VLbuumSCVuwo1MCe6dpVWKGUpARd9fCixu2N6puptXkl ufVx urf1 g288dq2tPG3HQfn/2n0 0bV ZzhjbVzNeWaXSQHFxx fG6V8Lt2jT3jJV1dRpUM/0Fi98X/fPJfpgQ4Hm3jpVU373lr51yjD9a8EW/avZnN4Tf/G6JCk7M1XvrMnX9WeM1M795dpfXq0nF23V26vztKfk4HO2f8zfpFeW79JDVx2rVbuKNGlITxVX1mhdXokefHeDvn3qcCUlJOhzfz7w3OvUUTkqrazR/W v0y8uHK VO4t08X3v69nvnKjr//WRdhbWXyT9wbRROnNsP339ofpRfDNTE1VWVauRfTO1raBc08b1k1T/2X/Tkx83bv/3X5yowb3SNbR3N6UnJ2pvaZWKKqr1y/ u1NKt /Xguxt07WnDG c2b14Yz7vtDN3/9jr1zUrT1oIyXXbcEM1avku3nDVaK3cU6dIH5qtPRopOGNFHZ43rpxufWKpbzx6jKaNylJ2Vokv/Ml 52d30/rqDP7PvmV0/QvLAHmmadeOpKq6o0YKNBTp WG/ll1QqJTHhoHPV9340VYN7ddOO/eVavHmfuiUnatq4ftq2r0xZqcktnttGgrXnKoiZPSZphKSlkhrKY ec 36Hd2yWKGmNpOmStklaJOly59zKJuucJ kG1ReTx0u61zl3/KG2nTpglBtw5T3aNOP8xmX7y6o06ZdvSJJ ev7h tapwzsaHSEWbGthNDo2t5cWbdqnn18wTgWlVfrO6SP0wYa9uvrRxXr uyfptmeXa/XuYh0xsLs25Jdqyc mqbK6Tr0yUlRdW6c653Tfm vqT1L7ZmpfWbV ev7h sMba/SnN9fpGyfn6psnD9PTi7dq3MDuqqiu01lH9FOCmbbtK1dJZY3GD6z/w/bSsp2aOXeDJOn8CQO0fX 5hmdnaPn2Qu0pqdTnjx6sD7fs08Ce6bpo4kBd89gSfff0EUpOTFBacqIumjRQJwXm/PzdpUdq9qe79dqKA1vj/nn18Ro7IEvPf7hdd876tHF5dmaqCkor23V/zVB/pVcAACAASURBVMTBPXTa6BxV1tbpr /U5/3tF47U/zy7TFccP7Txw12Scvt006a9B87teO9lk3TrM8s0tHc3rcsrOeC1m6fXf9Cv2FmovSVVKquqVXKiqbq2PtiInAwdN6yPXli6Xf/45nHaX1atV1fs0jGH9dLtzy1v3M4XjxmskX0zdc/stZo0pKfmx BJ/DdOztXD72 SJF127BC9uSpPeXE6MmNun24qr67V7qL4/PkQvMTAxQEA6Io2/ aCJc65yeHcR3uLyU8ljXMh7FtkZidKusM5d3bg e2S5Jy7u8k6f5X0tnPu34HnqyWd7pw7uG2/iebFZHFFtSbc8XqbeQb0SNMpI7P10db92l1YoeLKGn3j5FwlJyZo5twN v0XJ2rSkB6av6FAP/vPJ qblRqRE7KcrNSwDMndJyNFeyPcIggAAAAgMiJRTLarm6ukTyT1l9RmERekQZKaDo21TfWtj4daZ1CwOQ5VSErSzsIKPb3kwBvMG67mS9IPn/74gNcidWU/XHM7UUgCAAAA6Iz2FpPZklaa2UJJjdWNc 7CTuy7pZuWmrd8tmed hXNrpF0jSSl9B/ZuLxpF8qm3V4Ly6p101NL9eaqvHYH/tPlR mlZTv02ordGts/S6t2FeuyY4fo6yfm6vABWQfcp1FX51TnnP7 3kZt3Vemf36wpY0tAwAAAEBsaW8319NaWu6ce6fDOw5jN9dBo8a7rMt p4rqzwZSmXfbGRrYM72N70JzDe NtgYraU1tnQvc6P/ZwCcVNXVKSjClJCY0DnDQ9F7WruTaKcP19up8rd5drH7dU7W7qFJHDu6hyuo6bdxbqimjcnTU0J763WurJUkPX3WsvvHIosYb/hv87tIjdeesT WclJacwL1indBwgaglqUkJGtwrXevz2x4tEAAAIFpETTfXzhSNbVgkaZSZDZO0XdJlkr7SbJ0XJV1vZk ovgts4aEKSal NMHkJoXklycPoZDsgI4UkQ0SA8ViwzaSEhOU2cIokj27pSg9OVHlIRhOPdRaul916pgcdUtJ0tWnDlNyQoLGDshScmKC3lqdp9krd vOSyY0rvvEwi0aP6hH4wiTzd1 3uGNjxuK7 a N/WzVvamLetNfXHykIOWPbtkm04fk6M mamt/nzOOVXXOm3aW6p 3dMaR NssGVvmdblF2tPcZW OHmwyqpqtXJnkY7N7d3qNjftKdVhfbpp4cYCfXnmB/rr147RmH5ZenrJVm3fV66EBFNlTZ2umzJCSYmmIb3rRxxcubNIJ43I1ntr9 jGJz7SGzefph7pyY0j5j6zZJvuuPCIg/a3r7RKq3YVKyXJdNVDi5SanKifXXC4LjhyYOP/57Z9ZcrOTFVa8oGjZG7eW6ruacnqlZHS6s/TXDQPFjU8O0NTRufokXmbdGxuL33z5GF6fMEW/eyCcUpONA3LztCLH /QueMHqLq2Ttv3l v t9bp21OGa0jvbrr9ueX61inDtC6vROeM768Jd7yu8yb0150XT9CSzftUXVuneev36qwj qm0slbTDu rBDNt3VemXYUVmrd r755yjAlmPSP ZuVkpiglTuLdPfnJ2hDfqnyiiv07to9 vt7G3Xm2L6a06xXyL2XTdIdL67QRz8/S7uLKrR6V7F2F1Vod1GF mSmakROph56b6N2F1doYI903XHhEdpdVKHhORlyTkpPTtTWfWUymcykIb27aX1 iZ5dsk3/mL 5cZj4YdkZ ve3T9AJd382at9/vney7ntrnS47dogG9EhXVlqSXl6 U7uLKnRsbm 9vTpPt597uBITTV96YL7 7/Kj9Le5G/SFYwbruNzeenXFLp19RH/tK6vSs0u2aUROpk4dnS3n6offv 60EXpndb4uP36o/vTmWt145iilJiUqMcFUWVOr5IQE7SurUkVNnU6e8aaOG9ZbP5g2SvvLqvWHN9bo4auO1Suf7NRds1ZJko4b1lsLNxZE7s0VZd764emqrKnVU4u2aUz/TB1zWC t2FGk00f31c9f/EQnDO jzXvL9KXJg5WZlqQ GalKMOmlZTv10Psb9bevT9Yf3lijCyYM0Mh mXp52U4t2FCgV1fskiQ9fd2J tvcDXp95W599LPp2ltapYfe33jQyJeSlJ2ZoldunKJj75wtSVp757kqr67V4x9s0Yb8Eh03rLcWbCzQ2P5Z tapw/Wrl1bq7 9t1L2XTVJZVa3eX7dHV56U2ziEf1FFjYorqnXSiGxNGtJTzjkt2rRPo/tlKjUpUekpiSqprNGqnfUjSUrSjWeO0r1z1upXF4/Xzv3luv/t9Zp32xmatXynvn5irooqqjX517N14vA  sfVx2lXYITMJxdt1aBe6Zo5d4P 5 wxWrGjSEmJphOG99FlMz IxKEMqWHZGYecHkSSpozO0S8uPEJvr87TuAHd1bd7mqb /m1dctQgPf/RoaerkKSVvzxbBaVVOveed1VcWaMvTR6spxYfeKvUtMP7avan7e/99sHtZ2pnYbkWbizQ3a s0s3TR u/H /Q2rwSJSWYagIDSR05uIeWbQtuipNg3DJ9tP7fG2u07I6zdN9b6zS6b5ZuefpjfW7iwDanDZl98xTd/txyfW/qSP1x9lpNGtxDj84/cG7O5gNifeHowUpNTtC54/urzkmj 2XqrD/OVXFF/ef11DE5emt1fpt5H//W8fru4x/qzVtO05/eXKdH5m064PXfXnqkPtqyX9dOGa4VO4q0s7BcI/tmHjCqbltunj5af3hjTZvr/Pri8fr5C5/ohe dos/9 T39 LyxjZ/XrTlxeB9t31 uLQVljTnH9MvSRfe9f9C6U8fkqKK6TvdeNkl9u6cpr7hCx9158JQs0w7vp9mffjZgYr/uqYrE7KhttkyaWbFa7lZqqh/NtXundl4/Wus9qp8a5CHn3J1mdp3qN/5AYGqQP0s6R/VTg3zDObf4UNttGIBHkmZ9/1SNG9ipmAiza/6xWK9HcO6 Nb8 V1v3lWlgj3TVOqffvrpKt5w1RlmpSdqwp0Qj 2Yd9D0LNuxV74wUjep38GvoOnYWluvEu9/0HaPRqaOyNXVMX00c0kPHHNZ6kR9tSitr9K8FW3T1KcMiNs9sZU2tamqdMgLTkzSfDiWW7C2p1DG/nu07hhcrfnF24zEMtYaeNK0pLKtWVW2dcrJS5ZzTX95Zr0uPHqy 3dN0z w1qnP1J56RUlvn9P66PZoyOifk277vrXU6 4j Gtk3U4s2FeiLgcI1Gj38jWM1bkD3xul Sipr1C05UXtKKhunB6qurdNNTy7VurwSzfr qQf93jc/9k0vHJ4 JkdvBwqaW88eo88fPahx6iPnnIoqatQjPVk/eOIj/WfpjlYv/L68bKf690jTfz/eoQE90vT6yt3aWlCm2becpgSzA6ZOaqqiulaJCaYxP31F4wZ210s3nCpJenzBZm3bV66pY/rqqKE9tWhTgT7YUND4HswvrtQTC7fohjNHSZK27y/X/rIqjemXpaTEBG3IL9H8DXvVJyNFy7cX6r631ktq/cJ1g417SrVpT/1F6N /vlpDe3fT/PV79dpNUw5Yr7C8Wne8uEJ3BAr3aYf3U0pSQovTU7WloLRK89bv0cTBPdUtpf4i3NaCciUlmvKLK1t8/7 3do827ClRenJiixfcpfq5JgtKq5SeXH BZsf Co3om6GaWqerHl6oOy48QqP6ZiknK1WF5dXKL67Q6l0lOm1MjorKq9WzW7LSkxNb/cx4atFWJSSYjh7aUyWV9ftKSUpQUkKCjhras/H/YcnmAo3IyVTPbvUXtjfvLVVWWrJMUmKiNU4f1Nyuwgr1yUw55P nmUXHaK6xpqGY7JuVqoU/meY7Dg6hsLy6cW6jcLrvK0drTP8sjeybeeiVgVbMX79Xl/8tOq7aL/zxmQfMpYiuI5pbycPhjZumaHhOZos9OBB 0fp O1Th01ENP  cW05Tn4yUxttx2tpfw1gZSUEWS 1VU1snMwvb70BhebV2FVZoTH8umseTSBST4bm8FyVyslrv4ofo0drVuFB65BvH6vTApM5AZ3Si93fIUUiiq6BXCHwYkZOpypr623AOda6SkGBKaHHcyNAIV5HaoEd68kG3uwDtEdfF5PiBLd rhugSiSvNKWH EEbXES2dOZ669kTfEYCI PBn031HQBeWmpSom6eP1llH9PMdBYhKcX2GnZ3V/oE14F9WapLm3jpV828/I Tb7sxgQkBTk3N7 Y4gqX4QFiBePXzVsY2PewcxSBbC46UbTvEd4QDZmama9f1Tw7b9YdkZB7REfv/MURrbn/E3gJbEdctkYkJc18pxZd5tZ6h3RkrjiJt/v3Kyrn70kGMttRu1JEIl2MEDAARv6ti evY7JykjNfHQKyPsoq2gP2Vkn7AOrvjmLS3OiAegBXFZTPYKjIiUxI36MaP51C0nDO8Tsm0fMbC7jhxMl2cAiCXHHBYdvQAgJXSxK7L0ZgLaLy6LyYYiklHfYlcoP8dfDmNXGABAaN169hjfEdBMtJ1OXX3KcN8RAATEZX thvExaJmMXUl0UQZa9aXJg31HAMLmzMMZeTvqRNnp1AR6GwFRIy7P2Bvmzgz3MMoIn5SkBD129XGd3s6dl4wPQRoguvz20om I8Cz75w wneEsOFiYvSpq/OdAEC0iutPbFomY9upo3I6vY2sNOZMAhB/psVx692InAzfEdBMbbTMiSRpwY/P9B0BQBNxWUw2fORxzyROG935ghRo7qxxzDcG3 Lz71v3tCQGP4lCTafJ8K1f9zTfEQA0EZ/FZKCaTE7kD1JXdcMZI7VpxvnqkU7LJELvpumjfUcA4lIUNYChCf6WAmhNfBaTgbZJ7ruIfTlZqR36vsWb9oU4CQAg3MYPYmAVAIglcVltNVzZTKJlMuYd1rtbh75vaAe/D2gPn73w/nn18f52jigSn014M79 jO8IAIAgxGUx2YCWydjX0ftep46N38Ep0LV1tLUe8SVeu4MyaBoAxJbouaM6hBr yDL TuzrSOvy374 Oa5HOkTXNqZ/lu8IiAJ1cVZMfvWEoVq rdB3DLSDWfxezAAQvLgsJhswIlzsO/uI/np/3d6gvmc6I20izMzTSJq9utFqg3qujbP544f11oKNBRFM03EPXTVZA3uma2z/7r6joJ1M8drJGkBHxHU/UFomY9/XTjjMdwQgaiRwgQwBbbVMzvjCkZEL0gkv3XCKzhjbj0IyRvSNgi72Pz3/cN8RADQTl8Vkw2iunHjFPlqXEY2cp vy/D6gQVstk8OyMyKYpOP4Gx1bXr9pit764eleM5wzvr/X/QM4WHx2cw38jeXvFIBwqKvzs98fnsX8lqg3PCfTd4ROYzCp2NKzW4p6dktRenKiSqtqvWQY3IuR2oFoE5ctkw246gkgHOo8jT5x2XFDvewX0ad/jzRtmnG 7xhBS0lM0BPXnKBld5xFMRmjnv3uSb4jAIgicVlMNpzmUUvGh8s5gUaUGdgz3XcEoEVv3DTFd4Q23ThtlE4Y3kfdmQIkZnGPK4Cm4rKYbGiRTE6Myx vy GiAKJN74wU3xEASdLEIT0PeD6qX/ROHXPJUYP0rVOH Y4BAAihuKy2BvVK1/VTR rE4X18RwGAkBhEayhacP6E2BmQ5I9fnqTUpETfMQAAIRSXA/AkJZh ePYY3zEAIGTev 0M3xEQhZrOefq5iQM9JmnbvZdN8h0BMW5obwbfAaJRXLZMAkC4PXPdib4jADpxRH0PnJlfO0Z//NJEz2lad2EUF7qIDZOadekGEB0oJgGgAybn9taXJw JyL5G9Y39aSAQHuMH9dCmGefrrCP6KylKxwl4/rsnMUcqOi2BtxAQlaLzLw8AxIApo3Misp/Jub0jsh8gHI4a2st3BMSBK0/K9R0BQAsoJhH1uBiJaHX kQMisp87LhwXkf0AofbnrxzlOwLiBBclgOhEMYmY9p3TR j K472HQMIK0bARKw6Y2xf3xEAAGFEMYmod91pI1p9bUCPNJ03ITKtQ4APv7p4vO8IQIf845vHqVtKXA4aDwAIoJhE1BvCcODowjJTaZVE7HnzltMidk8xIu/iSYzOC6AexSRi2umj6UKF DawR7rvCEDQhucwAnE8O2lEtu8IAKIExSRiSr/uqQc8H9qHVkvEt9zsDN8RAAAAWuSlmDSz3mb2hpmtDfzb4hBdZvaQmeWZ2SeRzojo9OL1p/iOAERUnXO IwBB fkFjD4c9xhmHUCAr5bJ2yTNcc6NkjQn8Lwlj0g6J1KhEP36dU/zHQGIqDpqScSYb54yzHcEAECE ComL5L0aODxo5Iubmkl59xcSQWRCoXY9dS1J/qOAIRFn4wU3xGAdkugxapL4DADaOCrmOznnNspSYF/Oz2KipldY2aLzWxxfn5 pwMi h01tGfj4 OG9faYBAiPLxw9WGnJjOaK2PHaD6b4joAIyM5MPfRKALqEsE0AZWazJfVv4aWfhGN/zrmZkmZK0uTJk kY1gXcdckE3xGAsLrr88wxidjQOyNF3z19hEb1y/IdBRFw himfQFQL2zFpHNuWmuvmdluMxvgnNtpZgMk5YUrB LDtMP7KSXpwI41zccluXn6aN07Z20EUwGhNyInQ vzSyVJqUm0SiL6vfaDKRqRk6GkRAaI7yrM6OgKoJ6vT/4XJV0ZeHylpBc85UCMePDKybr/imPaXOf7Z47S rvOi1AioN5DV00O6fZ cv7hId0euqbvTR0RsX2N6Z9FIYmwunn6aN8RALTC16f/DEnTzWytpOmB5zKzgWY2q2ElM/u3pPmSxpjZNjO72ktaAGjFqL6h7daXmJCgnKxUDejByMXouC9PHuo7AhAyF00a6DsCgFaErZtrW5xzeyWd2cLyHZLOa/L88kjmQmzp0S3ZdwQg5Pp1T9XCH59JNzJ0SqTePuMHdY/MjgAAUclLMQl0xus3TVFJZY0G9Uz3HQUI UWNsf05OUfs6NWNqWsAoCujmETMGc1ogYgi3dOS1S0lUWVVtb6jABF3XC7TMiH8mg 4ByB6cMc8AHRSqFpnhmVnhGQ7AAAAkUAxCQCdNGloz5Bs56lrTwzJdgAgUh7/1vG IwDwiGISADrp95dO1AvfO7nT28nJSg1BGiByGCeq63rphlP0u0uPFG8BoGujmASATkpPSdTEIT112bFDfEcBIiolidOIrmr8oB764mQ 84Cujr8CABAid39 gjbNON93DCBirjwp13cEeJaVxjRdQFdGMQkAIcLckOhqUpMSfUeAZxMG9/AdAYBHFJMAEGIJ1JQA0Cmj 2WqTwbzmALRjmISAEJs8U n 44AADEtKy1ZmWn106EzzSQQvSgmASDEemek6Nopw33HAICIyO3TLeTbdI4SEogFFJMAEAY/Omes7wjowrh9F5H0zHdO0pPXnBDSbVJKArEhyXcAAIhHCdw4iTj3k/MO9x0BUSI7M1XZmaGdJ9c5MYclEANomQQAIM6Ee2ThB756jL5NV24084Npo0K6vW4p9W0eXJsDohfFJAB4dukxgzWoZ7rvGIgj3G8GH44f1idk23KS/nblZN169hgN7R36ezIBhAbdXAHAs19ceIQyUvk4RuzozZQNaEEoG8SdcxrUM13fmzoydBsFEHK0TAKAZ7QhIdTC2c314W8cq OG9Q7b9hG7jh7ay3cEABFGMQkAANpt6pi viMgSqUkhe60kp7aQGygmAQAzxhbAgAO5OizAcQEikkACJOzxvVr13qcMiHUwnWB4uGrjg3TloED0TIJxAaKSQAIk4lDevqOAITMxCE9NXUsXVwRGaHsMgsgfPhNBQAgzvTqFvrRVnNCPCk94lNKYudPLScO6an7vnJ0CNIACDeKSQAIkzDPGw 0Kj0lUZtmnN/h789uVjhmpSXp/31xYmdjoQuYdeMpnd7GTdNGaSBz7wIxgWISAAAcIDM18YDnT193onp0S/aUBrFkZN sg5ZdcfxQXTxp4EHLLz9uSCQiAQgjikkACBNjnFbEqOZjnzAYCjrjVxeN1y1njTlo fDsTF127MEFJW83IHZQTAIAgDZRTKKjjhzcQwkJpiG9u7X4 owvHBnhRABCiWISAMJk3MDuviMAHTa9ydQ2zPmHYHz4s mNj49qY1Rr7isHYh/FJACEyWmjc/TqD071HQMI2tFDe lH54xtfJ6ZmuQxDWJN74wU3XjmKA3okabbzzs86O/vkc79uUCsoJgEgDBiRELEors/P GA54f1yfCUBLHqpumjNf/2M5WWnHjolZs5emivMCQCEA4UkwDgWbcOnGwB7ZHewfdWWnJiYxfE4dkUkgCAllFMAoBHI3IylJDAjUMIj8m5tPAAAMKHYhIAADR68OuTfUcAAMQIL8WkmfU2szfMbG3g34MunZrZEDN7y8w NbMVZnajj6wAAHQl05qM4iox5x/C7xsn5/qOAKCDfLVM3iZpjnNulKQ5gefN1Ui6xTl3uKQTJH3PzMZFMCMAdNqhOrByoo5w6sz8kHS RqSM7pflOwKADvJVTF4k6dHA40clXdx8BefcTufch4HHxZI lTQoYgkBAAAQdly4AGKXr2Kyn3Nup1RfNErq29bKZpYr6ShJC8KeDABCyJiVGzHkC0cPPmiZ60zzJgAgroVtFmIzmy2pfwsv/STI7WRKelbSD5xzRW2sd42kayRp6NChwewCAPzhPB1hFOy1jJSkz76BCyGIFN5qQOwKWzHpnJvW2mtmttvMBjjndprZAEl5rayXrPpC8nHn3HOH2N9MSTMlafLkyZyeAYgKiZwlIcbxBxXhZnR0BWKWr26uL0q6MvD4SkkvNF/B6i J/l3Sp865P0QwGwCETHpKxyaNB0KBAXgQzU4Y3sd3BACd5KuYnCFpupmtlTQ98FxmNtDMZgXWOVnS1ySdYWZLA1/n YkLAB03oEea7whAh3HLJELlJ cdfsDz8YN6eEoCIFTC1s21Lc65vZLObGH5DknnBR6/Jy6MAogDbX2Q9e2eGrEcQDDooY1Q /aU4bpz1qcHv8B7DYhZvlomAQCS7r/iGN8RAAAAOoRiEgDCrK1RMXtnpEQwCbqazrQu9s2q7579/TNHhSgN0DIaJoHY5aWbKwAACL/O3O YnpKoTTPOD10YAEDcoWUSADy56qRc3xGAAzDYDgAgGBSTABBmrXU1vOPCIyIbBDgEBt2BD23dCgAgulFMAkCYcZ4EAADiEcUkAITZkF7dDlqWk8WUIIg dHOFD1xvA2IXxSQAhNn9Vxx90LL/Xn KhyToar44ebDvCMAh0XsDiF0UkwAQZj27HTz9R3IiZ08Iv4smDdLT153Y7vX70mKOCLlp2mjfEQCEAMUkAHhAb0JEo vPYE5JRMaN03ivAfGAYhIAgDgWTBt4ShKnBYg8urkCsYu/GgAQAcOyM3xHQBfVs1uy7wgAgDhFMQkAEXDJUYMOeM6omYiUkX2zfEcA2m3G5yf4jgAgCBSTABAB9OICgJZZk0/Iy44b6jEJgGBRTAKAB9wjBAAAYl2S7wAA0JVccOQAHT6gu7IzmYIBAADENopJAIiAhpbIw/p00/emjvQbBgA8 cWFR iYw3odsIyeGkDsopgEAABARFx5Uq7vCABCiHsmASACGubvS0lM9JwEAKLL4QO6 44AoINomQSACPj6ibkqLK/WtacN9x0FAKLK6H5MXwPEKopJAIiAtORE3Xr2WN8xAAAAQoZurgAAAACAoNEyCQAAtP6u83xHQBf20g2naOWOIt8xAASJYhIAACgxgfkZ4M/4QT00flAP3zEABIlurgAAAACAoFFMAgAAAACCRjEJAAAAAAgaxSQAAAAAIGgUkwAAAACAoFFMAgAAAACCRjEJAAAAAAgaxSQAAHHucxMH o4AAIhDFJMAAMS5P35pou8IAIA45KWYNLPeZvaGma0N/NurhXXSzGyhmX1sZivM7Bc sgIAEOuSEg/ cz/r 6d6SAIAiCe WiZvkzTHOTdK0pzA8 YqJZ3hnJsoaZKkc8zshAhmBAAgbo0b2N13BABAjPNVTF4k6dHA40clXdx8BVevJPA0OfDlIhMPAAAAANAWX8VkP fcTkkK/Nu3pZXMLNHMlkrKk/SGc25Baxs0s2vMbLGZLc7Pzw9LaAAAAABAvaRwbdjMZkvq38JLP2nvNpxztZImmVlPSc b2Xjn3CetrDtT0kxJmjx5Mi2YAAC04tRR2Qc8v fLkzwlAQDEsrAVk865aa29Zma7zWyAc26nmQ1QfctjW9vab2ZvSzpHUovFJAAA6JiLjxrkOwIAIAb56ub6oqQrA4 vlPRC8xXMLCfQIikzS5c0TdKqiCUEACBOjcjJlCS99oMp ssVR3tOAwCIVb6KyRmSppvZWknTA89lZgPNbFZgnQGS3jKzZZIWqf6eyZe8pAUAIMZ9buJASVKfjBT9 LzDJUlj mfp3AkDfMYCAMSwsHVzbYtzbq kM1tYvkPSeYHHyyQdFeFoAADEpf/93Dh1T0vS/37uCKUk bqWDACIJ16KSQAAEFnZmam685IJvmMAAOIIlyYBAAAAAEGjmAQAAAAABI1iEgAAAAAQNIpJAAAAAEDQKCYBAAAAAEGjmAQAAAAABI1iEgAAAAAQNIpJAAAAAEDQKCYBAAAAAEGjmAQAAAAABI1iEgAAAAAQNIpJAAAAAEDQKCYBAAAAAEEz55zvDCFnZsWSVvvOgVZlS9rjOwTaxDGKbhyf6Mcxin4co jG8Yl HKPoN8Y5lxXOHSSFc MerXbOTfYdAi0zs8Ucn jGMYpuHJ/oxzGKfhyj6MbxiX4co hnZovDvQ 6uQIAAAAAgkYxCQAAAAAIWrwWkzN9B0CbOD7Rj2MU3Tg 0Y9jFP04RtGN4xP9OEbRL zHKC4H4AEAAAAAhFe8QKjqIQAACIJJREFUtkwCAAAAAMIoropJMzvHzFab2Tozu813nnhmZkPM7C0z 9TMVpjZjYHlvc3sDTNbG/i3V5PvuT1wbFab2dlNlh9jZssDr/2fmVlgeaqZPRlYvsDMciP9c8Y6M0s0s4/M7KXAc45PFDGznmb2jJmtCvwuncgxii5mdlPgM 4TM/u3maVxjPwys4fMLM/MPmmyLCLHxMyuDOxjrZldGZmfOLa0cnx F/icW2Zmz5tZzyavcXwirKVj1OS1H5qZM7PsJss4RhHW2jEysxsCx2GFmf22yXJ/x8g5FxdfkhIlrZc0XFKKpI8ljfOdK16/JA2QdHTgcZakNZLG/f/27j/W6rqO4/jztW4CV6aCjDKgwAI2dU4NCNMswZEZg2r gcsF/Vgt6w9zq2zXudr8Q5vLtlySwzULfxMT5mpCOXI5BYyhoGjcgOQqJsxUygWS7/74vE8eT/deOHDvPV/OfT227873 /6e7/n14u58P fz X4Afgxcm/VrgZty/YzMZAQwJbN6T 7bAJwPCPgd8JmsXwUszfVFwH2tft/H2wJcA9wNPJTbzqdCC3An8LVcPwE4xRlVZwEmADuBUbl9P7DEGbU8l4uA84CtdbVBzwQYC zI2zG5PqbVn0fVlj7ymQd05PpNzqd6GWV9EvAw8DdgnDOqVkbAxcDvgRG5Pb4KGbVTz QsoDsidkTEQeBeYGGLX1Pbiog9EbEp1/cD2ygnXgspJ8jk7edyfSFwb0QciIidQDcwS9JpwEkR8XiUf8W/ajim9lgrgLm1X1Ts8CRNBD4LLKsrO5 KkHQS5cviDoCIOBgRr GMqqYDGCWpA gEXsIZtVREPAq82lAeikw DayNiFcj4h/AWuDSgX Hx7fe8omINRFxKDefACbmuvNpgT7 hgBuAb4H1E o4oxaoI MvgncGBEH8j6vZL2lGbVTY3ICsLtuuydrNsiya/xcYD3wvojYA6XBCYzPu/WVz4Rcb6y/65j8EnodOHUw3kOb inlS HtuprzqY7Tgb3AL1WGIi TdCLOqDIi4kXgZuAFYA/wekSswRlV0VBk4vOMgfEVSg8JOJ/KkLQAeDEinmrY5YyqYxrwiRyW kdJM7Pe0ozaqTHZ2y 5nqp2kEkaDfwGuDoi3ujvrr3Uop96f8fYYUiaD7wSEX8 0kN6qTmfwdVBGcJyW0ScC/yLMjyvL85oiKlcd7eQMmzoA8CJkq7s75Beas6otQYyE2d1jCR1AYeAu2qlXu7mfIaYpE6gC7i t9291JxRa3RQhp7OBr4L3J 9iS3NqJ0akz2Usd41EynDkWyQSHovpSF5V0SszPLfs1udvK11wfeVTw/vDHepr7/rmBxidjK9D8uw/3cBsEDSLsqQ7zmSluN8qqQH6ImI9bm9gtK4dEbVcQmwMyL2RsRbwErg4zijKhqKTHyecQxyIo/5wBdzyB04n6r4MOVHs6fyvGEisEnS 3FGVdIDrIxiA2Xk2ThanFE7NSY3AlMlTZF0AuVi0tUtfk1tK38JuQPYFhE/qdu1GqjN/LQYWFVXX5SzR00BpgIbcjjSfkmz8zG/1HBM7bEuBx6p wKyfkTEDyJiYkRMpvwtPBIRV J8KiMiXgZ2S5qepbnAszijKnkBmC2pMz/buZTrw51R9QxFJg8D8ySNyV7reVmzw5B0KfB9YEFEvFm3y/lUQERsiYjxETE5zxt6KJMsvowzqpIHgTkAkqZRJu7bR6szigrMWDRQC3AZZVbRvwJdrX497bwAF1K6vZ8GNudyGWW89R A7Xk7tu6YrszmeXI2qazPALbmvlsBZX0k8ADlQuINwOmtft/H4wJ8indmc3U FVqAc4An8 /oQcrwFWdUoQX4EfBcfr6/psyW54xam8k9lGtY36Kc9H51qDKhXO/XncuXW/1ZVHHpI59uynVYtfOFpc6nWhk17N9FzubqjKqTEaXxuDw/803AnCpkVHtAMzMzMzMzsyPWTsNczczMzMzMbIi4MWlmZmZmZmZNc2PSzMzMzMzMmubGpJmZmZmZmTXNjUkzMzMzMzNrmhuTZmY27EnqkvSMpKclbZb0sUF8rnWSZgzW45uZmQ2Vjla/ADMzs1aSdD4wn/KfdB QNI7y/3mZmZlZP9wzaWZmw91pwL6IOAAQEfsi4iVJ10vaKGmrpNslCf7Xs3iLpEclbZM0U9JKSdsl3ZD3mSzpOUl3Zm/nCkmdjU8saZ6kxyVtkvSApNFZv1HSs3nszUP4WZiZmR0xNybNzGy4WwNMkvQXST X9Mms3xoRMyPiLGAUpfey5mBEXAQsBVYB3wLOApZIOjXvMx24PSLOBt4Arqp/0uwBvQ64JCLOA54ErpE0Fvg8cGYee8MgvGczM7Nj5sakmZkNaxHxT CjwNeBvcB9kpYAF0taL2kLMAc4s 6w1Xm7BXgmIvZkz YOYFLu2x0Rj X6cuDChqeeDZwBPCZpM7AY BCl4flvYJmkLwBvDtibNTMzG0C ZtLMzIa9iPgPsA5Yl43HbwBnAzMiYrekHwIj6w45kLdv163XtmvfrdH4NA3bAtZGxBWNr0fSLGAusAj4NqUxa2ZmVinumTQzs2FN0nRJU tK5wDP5/q vI7x8qN46A/m5D4AVwB/atj/BHCBpI/k6 iUNC2f7 SI C1wdb4eMzOzynHPpJmZDXejgZ9JOgU4BHRThry RhnGugvYeBSPuw1YLOkXwHbgtvqdEbE3h9PeI2lElq8D9gOrJI2k9F5 5yie28zMbNAponHUjZmZmR0LSZOBh3LyHjMzs7bkYa5mZmZmZmbWNPdMmpmZmZmZWdPcM2lmZmZmZmZNc2PSzMzMzMzMmubGpJmZmZmZmTXNjUkzMzMzMzNrmhuTZmZmZmZm1jQ3Js3MzMzMzKxp/wVIhqSuA02QSQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 1080x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "print(\"Sample rate =\\t\"   str(sr)   \" samples/s\")\n",
    "print(\"Duration t =\\t\"   str(len(myrec) / sr)   \" s\")\n",
    "print(\"Average intensity =\",np.mean(np.abs(myrec)))\n",
    "\n",
    "# Play it\n",
    "play = False\n",
    "if play:\n",
    "    sd.play(myrec, sr)\n",
    "    \n",
    "# Plot\n",
    "plt.figure(figsize=(15, 6))\n",
    "plt.plot(myrec)\n",
    "plt.xlabel(\"Samples\")\n",
    "plt.ylabel(\"Intensity\")\n",
    "plt.xlim(0, len(myrec))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2) Load Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Load Model\n",
    "from tensorflow.keras.models import load_model\n",
    "model_name=\"2020-07-25_13-50_DSConvModel\"\n",
    "model = load_model('models/' model_name '.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3) Sliding Window\n",
    "A 1 second window will slide left to right\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of windows: 64\n",
      "\n",
      "\n",
      "Window 0 Silence\n",
      "Window 1 Silence\n",
      "Window 2 Silence\n",
      "Window 3 Silence\n",
      "Window 4 yes\n",
      "Window 5 yes\n",
      "Window 6 yes\n",
      "Window 7 yes\n",
      "Window 8 yes\n",
      "Window 9 yes\n",
      "Window 10 Silence\n",
      "Window 11 Silence\n",
      "Window 12 Silence\n",
      "Window 13 Silence\n",
      "Window 14 Silence\n",
      "Window 15 Silence\n",
      "Window 16 Silence\n",
      "Window 17 Silence\n",
      "Window 18 no\n",
      "Window 19 no\n",
      "Window 20 no\n",
      "Window 21 no\n",
      "Window 22 no\n",
      "Window 23 no\n",
      "Window 24 no\n",
      "Window 25 up\n",
      "Window 26 Silence\n",
      "Window 27 Silence\n",
      "Window 28 Silence\n",
      "Window 29 Silence\n",
      "Window 30 Silence\n",
      "Window 31 yes\n",
      "Window 32 down\n",
      "Window 33 down\n",
      "Window 34 down\n",
      "Window 35 down\n",
      "Window 36 down\n",
      "Window 37 down\n",
      "Window 38 on\n",
      "Window 39 up\n",
      "Window 40 yes\n",
      "Window 41 Silence\n",
      "Window 42 Silence\n",
      "Window 43 Silence\n",
      "Window 44 left\n",
      "Window 45 left\n",
      "Window 46 left\n",
      "Window 47 left\n",
      "Window 48 left\n",
      "Window 49 left\n",
      "Window 50 left\n",
      "Window 51 left\n",
      "Window 52 left\n",
      "Window 53 Silence\n",
      "Window 54 Silence\n",
      "Window 55 Silence\n",
      "Window 56 Silence\n",
      "Window 57 Silence\n",
      "Window 58 Silence\n",
      "Window 59 Silence\n",
      "Window 60 Silence\n",
      "Window 61 Silence\n",
      "Window 62 Silence\n",
      "Window 63 Silence\n",
      "\n",
      "Average prediction speed per window (ms): 129.74\n"
     ]
    }
   ],
   "source": [
    "###################### PARAMETERS ##########################################\n",
    "\n",
    "ws=7 #windows per second >=4\n",
    "\n",
    "#############################################################################\n",
    "\n",
    "if len(myrec.shape)<2: myrec=myrec[...,np.newaxis] # for the 1 sec audios\n",
    "\n",
    "step=int(sr/ws)\n",
    "nwin=int((len(myrec)-sr)/step) 1 #number of windows\n",
    "windows=np.empty((nwin,sr,1))\n",
    "predictions=np.empty((nwin,1))\n",
    "\n",
    "print(\"Number of windows:\",nwin)\n",
    "#print(\"Step:\",step)\n",
    "print(\"\\n\")\n",
    "#print(windows.shape)\n",
    "#print(predictions.shape)\n",
    "\n",
    "#divide in windows\n",
    "for i in range(nwin):\n",
    "    start=int(i*step)\n",
    "    end=int(start sr)\n",
    "    windows[i]=myrec[start:end]\n",
    "\n",
    "    \n",
    "\n",
    "#prediction speed measure\n",
    "import time\n",
    "sStarts=[]\n",
    "sEnds=[]\n",
    "\n",
    "for i in range(nwin):\n",
    "    start=int(i*step)\n",
    "\n",
    "    #SILENCE FILTER\n",
    "    if np.mean(np.abs(windows[i]))<5*(10**-3):\n",
    "        predictions[i]=-1 #Silence\n",
    "        #plt.text(start sr/4, 0, \"SILENCE \" str(i), color='green')\n",
    "        print(\"Window\",i,\"Silence\")\n",
    "    else:\n",
    "        sStarts.append(time.time())\n",
    "        \n",
    "        a=windows[i]\n",
    "        \n",
    "        #preprocessing:         \n",
    "        a=np.squeeze(a) #remove dimensions of length 1 (the last dimension)\n",
    "        a=a[np.newaxis,...] #add dimension 1 at the beginning to indicate 1 example\n",
    "        #a=LoadAndPreprocessDataset.MFCC(a,n_mfcc=40)\n",
    "        a=LoadAndPreprocessDataset.melspect(a)\n",
    "        a=a[...,np.newaxis] #add dimension 1 at the end for CNN\n",
    "        \n",
    "        #no preprocessing:\n",
    "        #a=a[np.newaxis,...]\n",
    "        \n",
    "        predictions[i]=np.argmax(model.predict(a),1)\n",
    "        \n",
    "        p=categories[int(predictions[i])]\n",
    "        #plt.text(start sr/4, 0.1, \"PRED \" str(i), color='blue')\n",
    "        print(\"Window\",i,p)\n",
    "    \n",
    "        sEnds.append(time.time())\n",
    "    \n",
    "#plt.show()\n",
    "\n",
    "\n",
    "#Prediction Speed measure\n",
    "elapseds=[]\n",
    "for i in range(len(sStarts)):\n",
    "    elapseds.append(sEnds[i]-sStarts[i])\n",
    "\n",
    "\n",
    "print(\"\\nAverage prediction speed per window (ms):\",str(round(np.mean(elapseds)*1000,2)))\n",
    "#print(\"Total elapsed time (ms):\", str(round(np.mean(elapseds)*1000*len(elapseds),2)))\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Fusion Rule"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Window 0 Silence\n",
      "Window 1 Silence\n",
      "Window 2 Silence\n",
      "Window 3 Silence\n",
      "Window 4 yes\n",
      "Window 5 yes\n",
      "Window 6 yes\n",
      "Window 7 yes\n",
      "Window 8 yes\n",
      "Window 9 yes\n",
      "Window 10 Silence\n",
      "Window 11 Silence\n",
      "Window 12 Silence\n",
      "Window 13 Silence\n",
      "Window 14 Silence\n",
      "Window 15 Silence\n",
      "Window 16 Silence\n",
      "Window 17 Silence\n",
      "Window 18 no\n",
      "Window 19 no\n",
      "Window 20 no\n",
      "Window 21 no\n",
      "Window 22 no\n",
      "Window 23 no\n",
      "Window 24 no\n",
      "Window 25 no\n",
      "Window 26 Silence\n",
      "Window 27 Silence\n",
      "Window 28 Silence\n",
      "Window 29 Silence\n",
      "Window 30 Silence\n",
      "Window 31 down\n",
      "Window 32 down\n",
      "Window 33 down\n",
      "Window 34 down\n",
      "Window 35 down\n",
      "Window 36 down\n",
      "Window 37 down\n",
      "Window 38 down\n",
      "Window 39 down\n",
      "Window 40 down\n",
      "Window 41 Silence\n",
      "Window 42 Silence\n",
      "Window 43 Silence\n",
      "Window 44 left\n",
      "Window 45 left\n",
      "Window 46 left\n",
      "Window 47 left\n",
      "Window 48 left\n",
      "Window 49 left\n",
      "Window 50 left\n",
      "Window 51 left\n",
      "Window 52 left\n",
      "Window 53 Silence\n",
      "Window 54 Silence\n",
      "Window 55 Silence\n",
      "Window 56 Silence\n",
      "Window 57 Silence\n",
      "Window 58 Silence\n",
      "Window 59 Silence\n",
      "Window 60 Silence\n",
      "Window 61 Silence\n",
      "Window 62 Silence\n",
      "Window 63 Silence\n"
     ]
    }
   ],
   "source": [
    "#majority\n",
    "'''\n",
    "DFA\n",
    "         -1           any = No silence\n",
    "         ↓ ↑   any   ↓ ↑\n",
    "    --->(S1)--------->(S2)\n",
    "          ↑            |\n",
    "          --------------\n",
    "                -1 = Silence\n",
    "'''\n",
    "newPredictions=[]\n",
    "predictions=list(predictions)\n",
    "predictions =[-1]*ws #add silence at the end (will be removed later)\n",
    "\n",
    "s1=True #true=s1, false=s2\n",
    "i=0\n",
    "start=0\n",
    "while i<len(predictions):\n",
    "    if s1: #state 1\n",
    "        if predictions[i]!=-1: #not silence\n",
    "            start=i\n",
    "            s1=False\n",
    "        else: #silence\n",
    "            newPredictions.append(-1)\n",
    "        i =1\n",
    "    else: #state 2\n",
    "        if predictions[i]!=-1: #not silence\n",
    "            pass\n",
    "        else: #silence\n",
    "            end=i\n",
    "            a=predictions[start:end]\n",
    "            #calculate most common number in the list\n",
    "            ht={}\n",
    "            freq=0\n",
    "            most_freq=-1\n",
    "            for e in a:\n",
    "                e=int(e)\n",
    "                if e in ht:\n",
    "                    ht[e] =1\n",
    "                    if ht[e]>freq:\n",
    "                        most_freq=e\n",
    "                        freq=ht[e]\n",
    "                else: ht[e]=0\n",
    "            newPredictions=newPredictions [most_freq]*(end-start)\n",
    "            #append silence\n",
    "            newPredictions.append(-1)\n",
    "            s1=True\n",
    "        i =1\n",
    "#print(newPredictions)\n",
    "\n",
    "# remove the silence added before\n",
    "del newPredictions[-ws:]\n",
    "\n",
    "for i in range(len(newPredictions)):\n",
    "    #if newPredictions[i]==-1: print(i,\"Read:\",predictions[i],newPredictions[i],\"Silence\")\n",
    "    #else: print(i,\"Read:\",predictions[i],newPredictions[i],categories[newPredictions[i]])\n",
    "    if newPredictions[i]==-1: print(\"Window\",i,\"Silence\")\n",
    "    else: print(\"Window\",i,categories[newPredictions[i]])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA5MAAAE9CAYAAACWZjY AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1 /AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy WH4yJAAAgAElEQVR4nOzdd3iUVf7 8ffJpJNK6DX0Jk1QQQU7NixY1l5XkVVWdNVVV11xV9f63fqzrLr2tvaCiLpYAJHee5MSCJ2EhNSZOb8/EkJLz8ycSXK/riuXU06e504MM/N5TjPWWkRERERERERqIsJ1ABEREREREal/VEyKiIiIiIhIjamYFBERERERkRpTMSkiIiIiIiI1pmJSREREREREakzFpIiIiIiIiNRYpOsAwdCsWTObnp7uOoaIiIiIiIgTc fO3WmtbR7MczTIYjI9PZ05c a4jiEiIiIiIuKEMWZDsM hYa4iIiIiIiJSYyomRUREREREpMZUTIqIiIiIiEiNNcg5kyIiIiIi0jAUFxeTkZFBQUGB6yhhKTY2lnbt2hEVFRXyc6uYFBERERGRsJWRkUFiYiLp6ekYY1zHCSvWWnbt2kVGRgadOnUK fk1zFVERERERMJWQUEBaWlpKiTLYYwhLS3NWa tikkREREREQlrKiQr5vJ347SYNMacZYxZaYxZY4y5r5J2xxhjfMaYS0KZT0REREREBOCxxx6jT58 9OvXjwEDBjBz5kxuuukmli1bBpTsdb9z507HKUPL2ZxJY4wHeBY4A8gAZhtjPrfWLiun3ZPA16FPKSIiIiIijd3PP//MhAkTmDdvHjExMezcuZOioiJefvll19GcctkzeSywxlq7zlpbBLwHXFBOu98CHwHbQxlORESkIftl4UrWLVjhOoaISL2QmZlJs2bNiImJAaBZs2a0adOGk08 mTlz5hzR/q233uLYY49lwIAB3HLLLfh8PgASEhJ44IEH6N /P0OGDGHbtm0AbNu2jVGjRtG/f3/69 /P9OnTKz1OuHBZTLYFNh10P6P0sTLGmLbAKOCFEOYSERFp0JZNnUPBZRdSePkofN7w mAiIhKORowYwaZNm jevTu33norP/74Y4Vtly9fzn// 19  uknFixYgMfj4e233wZg3759DBkyhIULFzJ8 HBeeuklAG6//XZOOukkFi5cyLx58 jTp0 lxwkXLrcGKW mqD3s/t Be621vqomlhpjRgOjATp06BCQgCIiIg2Rufmasttf3fc4I5950GEaEZHqe SLpSzbsjegx zdJomHz tTaZuEhATmzp3L1KlT f7777nssst44oknym07efJk5s6dyzHHHANAfn4 LVq0ACA6OpqRI0cCMGjQIL799lsAvvvuO9544w0APB4PycnJvPnmmxUeJ1y4LCYzgPYH3W8HbDmszWDgvdJCshlwjjHGa6399PCDWWtfBF4EGDx48OFFqYiIiJTDrtJQVxGR6vB4PJx88smcfPLJ9O3bl9dff73cdtZarrvuOh5//PEjnouKiipbfdXj8eD1eis8X2XHCRcui8nZQDdjTCdgM3A5cOXBDay1ZTtvGmNeAyaUV0iKiIiIiEjDV1UPYrCsXLmSiIgIunXrBsCCBQvo2LEjS5YsOaLtaaedxgUXXMCdd95JixYt2L17Nzk5OXTs2LHC45922mk8//zz3HHHHfh8Pvbt21er44SaszmT1lovMJaSVVqXA 9ba5caY8YYY8a4yiUiIiIiInKw3NxcrrvuOnr37k2/fv1YtmwZ48ePL7dt7969efTRRxkxYgT9 vXjjDPOIDMzs9Lj/ Mf/ D777 nb9  DBo0iKVLl9bqOKFmrG14I0IHDx5sy1tVSURERGB5z15lt9d0H8R5n7/lMI2ISOWWL19Or169qm7YiJX3OzLGzLXWDg7meV2u5ioiIiIiIiL1lIpJERERERERqTEVkyIiIiIiIlJjKiZFRERERESkxlRMioiIiIiISI2pmBQREREREZEaUzEpIiIiIiIiNaZiUkRERERERGos0nUAERERERGRcPbQQw/RrFkzxo0bB8ADDzxAy5YtKSws5P3336ewsJBRo0bxyCOPsG/fPn71q1 RkZGBz fjoYce4rLLLnP8EwSHeiZFREREREQq8etf/5rXX38dAL/fz3vvvUfLli1ZvXo1s2bNYsGCBcydO5cpU6YwadIk2rRpw8KFC1myZAlnnXWW4/TBo55JERERERGpH766D7YuDuwxW/WFs5 otEl6ejppaWnMnz fbdu2MXDgQGbPns0333zDwIEDAcjNzWX16tUMGzaMu   m3vvvZeRI0cybNiwwOYNIyomRUREREREqnDTTTfx2muvsXXrVm688UYmT57M/fffzy233HJE27lz5zJx4kTuv/9 RowYwR// EcHiYNPxaSIiEgjsnd3tusIIiK1V0UPYjCNGjWKP/7xjxQXF/POO 8QGRnJQw89xFVXXUVCQgKbN28mKioKr9dL06ZNufrqq0lISOC1115zljnYVEyKiIg0ItMvvpqOhzxiHSUREalfoqOjOeWUU0hJScHj8TBixAiWL1/O0KFDAUhISOCtt95izZo13HPPPURERBAVFcXzzz/vOHnwqJgUERFpRDpmrnEdQUSkXvL7/cyYMYMPPvig7LFx48aVrfC6X5cuXTjzzDNDHc8JreYqIiIiIiJSiWXLltG1a1dOO 00unXr5jpO2FDPpIiIiIiISCV69 7NunXrXMcIO qZFBERERERkRpTMSkiIiIiIiI1pmJSREREREREakzFpIiIiIiIiNSYikkREZFGzu/3u44gIiL1kIpJERGRRqzrqnl8cdNdrmOIiEg9pGJSRESkkes fZLrCCIiYW39 vX06tWLm2  mT59 jBixAjy8/NZsGABQ4YMoV /fowaNYo9e/a4jhpSKiZFRERERESqsHr1am677TaWLl1KSkoKH330Eddeey1PPvkkixYtom/fvjzyyCOuY4ZUpOsAIiIi4t6auUvoOugo1zFERCr15KwnWbF7RUCP2bNpT 499t4q23Xq1IkBAwYAMGjQINauXUtWVhYnnXQSANdddx2XXnppQLOFO6c9k8aYs4wxK40xa4wx95Xz/AXGmEXGmAXGmDnGmBNd5BQREWno8rP3uY4gIhLWYmJiym57PB6ysrIcpgkPznomjTEe4FngDCADmG2M dxau ygZpOBz6211hjTD3gf6Bn6tCIiIiIi4lp1ehBDJTk5mdTUVKZOncqwYcN48803y3opGwuXw1yPBdZYa9cBGGPeAy4AyopJa23uQe2bADakCUVERBoJe9uv2fjRZ3To3cV1FBGReuP1119nzJgx5OXl0blzZ1599VXXkULKZTHZFth00P0M4LjDGxljRgGPAy2Ac0MTTUREpHGJsj7mP/UvOrz2d9dRRETCTnp6OkuWLCm7f/fdd5fdnjFjhotIYcHlnElTzmNH9Dxaaz x1vYELgT XOHBjBldOq9yzo4dOwIYU0REpLHQACAREak l8VkBtD oPvtgC0VNbbWTgG6GGOaVfD8i9bawdbawc2bNw9sUhERkUbA5OW5jiAiIvWIy2JyNtDNGNPJGBMNXA58fnADY0xXY4wpvX00EA3sCnlSERGRRqDboml4i72uY4iISD3hrJi01nqBscDXwHLgfWvtUmPMGGPMmNJmFwNLjDELKFn59TJrrcbgiIiIBMmXN97hOoKIiNQTLhfgwVo7EZh42GMvHHT7SeDJUOcSERFprDrMm o6goiI1BMuh7mKiIhImDFahEdERKpJxaSIiIgcoNkkIiJVGj9 PM8884zrGM6pmBQREREREZEaUzEpIiIiZcrbBFokFBYs3cGyVVq0X8LXY489Ro8ePTj99NNZuXIlAAsWLGDIkCH069ePUaNGsWfPHrZv386gQYMAWLhwIcYYNm7cCECXLl3Iy8vj uuv5/bbb f444 nc fOfPjhh85 rrpQMSkiItJI Ly KttozqS48tO/FvP9Xxe6jiFSrrlz5/Lee 8xf/58Pv74Y2bPng3Atddey5NPPsmiRYvo27cvjzzyCC1atKCgoIC9e/cydepUBg8ezNSpU9mwYQMtWrQgPj4egMzMTKZNm8aECRO47777XP54teZ0NVcREREJnSXfzyDadQgRkTrY pe/ULh8RUCPGdOrJ63 8IdK20ydOpVRo0aVFYLnn38  /btIysri5NOOgmA6667jksvvRSA448/np9  okpU6bwhz/8gUmTJmGtZdiwYWXHvPDCC4mIiKB3795s27YtoD9TqKhnUkREpJHw /1VN1LHpIhIuYyp/kSAYcOGlfVGXnDBBSxcuJBp06YxfPjwsjYxMTFlt209XfxMPZMiIiJSJsr6mPb6J5x43SjXUUREjlBVD2KwDB8 nOuvv5777rsPr9fLF198wS233EJqaipTp05l2LBhvPnmm2W9lMOHD fBBx9k PDhRERE0LRpUyZOnMjjjz/uJH wqJgUERGRQ6Q9/gdQMSkiUuboo4/msssuY8CAAXTs2LFsuOrrr7/OmDFjyMvLo3Pnzrz66qsApKenA5T1RJ544olkZGSQmprqJH wmPrapVqZwYMH2zlz5riOISIiElbmfz2V2HGjq9W214rlQU4jcqhnx3wHwG0vnOo4iYSb5cuX06tXL9cxwlp5vyNjzFxr7eBgnldzJkVERBoLf8O7gCwiIu5omKuIiEgj8Nn14wBDd9dBRESkwVAxKSIi0gh0n/GN6wgiItLAaJiriIiIiIiI1JiKSREREREREakxFZMiIiIiEjbyC7yuI4hINamYFBEREZGw8e/n5rmOIHKEhISEKtv885//pFevXlx11VV8 umnLFu2LATJ3FIxKSIiIiJhoyCn2HUEkVp57rnnmDhxIm //baKSRERERERETnU008/zTHHHEO/fv14 OGHARgzZgzr1q3j/PPP57HHHuPzzz/nnnvuYcCAAaxdu9Zx4uDR1iAiIiIiIiLV8M0337B69WpmzZqFtZbzzz fKVOm8MILLzBp0iS //57mjVrxurVqxk5ciSXXHKJ68hBpWJSRERERETqhanvr2LnptyAHrNZ wSG/ap7tdp 8803fPPNNwwcOBCA3NxcVq9ezfDhwwOaqb5QMSkiIiIiIlIN1lruv/9 brnlFtdRwoKKSRERERERqReq24MYLGeeeSYPPfQQV111FQkJCWzevJmoqChatGhxSLvExERycnIcpQwdLcAjIiIiIiH175cW8MEnK13HEKmxESNGcOWVVzJ06FD69u3LJZdcUm7RePnll/P0008zcOBALcAjIiIiIhIIhUVevHN3sx1gVA/XcUSqJTf3wDzNcePGMW7cuCParF /vuz2CSec0Ci2BlExKSIiIiIhsXNPPv 9/2fXMUQkQJwOczXGnGWMWWmMWWOMua c568yxiwq/ZpujOnvIqeIiIiI1N3WbXlVtknMLAxBEhEJBGfFpDHGAzwLnA30Bq4wxvQ rNkvwEnW2n7An4EXQ5tSREREREREyuOyZ/JYYI21dp21tgh4D7jg4AbW2unW2j2ld2cA7UKcUUQk4N5fPI1FW9e7jiEi4ty7HywnO6eQYq/fdRQJc9Za1xHClsvfjcs5k22BTQfdzwCOq6T9r4GvKnrSGDMaGA3QoUOHQOQTEQmKP8/7DdZGsOT6ha6jSCOxdv5y1xFEyrV7cibPL91NXHK0FvKQCsXGxrJr1y7S0tIwxriOE1astezatYvY2Fgn53f577a8v4Ryy2pjzCmUFJMnVnQwa 2LlA6DHTx4sC5diEhYM0ZX4SV0tt94PSmuQ4hUwJ9VSO6 Yv2NSoXatWtHRkYGO3bscB0lLMXGxtKunZsBnC6LyQyg/UH32wFbDm9kjOkHvAycba3dFaJsIiIiDUZsUYHrCCIAWHvkhbTkAsiKchBG6o2oqCg6derkOoaUw WcydlAN2NMJ2NMNHA58PnBDYwxHYCPgWustascZBQREREREZFyOOuZtNZ6jTFjga8BD/CKtXapMWZM6fMvAH8E0oDnSsdHe621g11lFhERERERkRJO5zpbaycCEw977IWDbt8E3BTqXCIiofDDuiWc3Pko1zFEREREasXlMFcRkUbttz/8mqXbNlXdUKSOrBY/FBGRIFAxKSLiiieP0V/93nUKkXL5/VpxWEInJUd/byL1kYpJEZEQ2ldYeMh9v/Vx0uvXceZbtzpKJI2BqcWGWVvXZgQ iIiINCgqJkVEQmjkf39zyP08/w52M48tvqms273NUSqRI6367idmfjDJdQyppz78bCV//c1kcvOKXUcRkSBSMSkiEkI77exD7vsjd5bdvnXin0IdRxqJ2syZbPm3P5H00J0snz4v8IGkwVv77WZirCFjS47rKCISRComRUTChKUWYxFFgixvV5brCFLPbN2 jwRvye2P/7WAv4z7zm0gEQkaFZMiImHCoCU3JfAmPvgMcd4i1zGkEfnvH2eU3U4uLPkq9pYssDPxg1XVOobPqwV5ROoDp/tMioiISHB1 vA/riNII5KdU0hkORfG/v7UTIp2FZCyr3ojMD74dBWnnNSels2bBDqiiASQeiZFREREJCDefXdZuY/Hb8yvdiEJsOt/W3j5qdlVNxQRp1RMioiISIX2rNvgOoLUJwGc u0p0FBXkXCnYlJEJEQysndX/nzx9BAlEam ts8/5TqCiIiEKRWTIiIhMitjZaXPmwhviJKIiARH0fzKL5qJSMOiYlJEREQqpb0mpbo8WpVapFFRMSkiIiKVu/Eq1wlERCQMqZgUERERkbCTWAw//pzhOoaIVELFpIhIiNhALnMoEmLrF1Vvs3lpvIq9gV99dcnr rsTCWcqJkVERKRKS 5/2HUECXPPjf3edQQRCTEVkyIiIlIl4y12HUHCWLHXT6QW3xFpdFRMioiEiNEHLanHOm9Y6jqCiIiEGRWTIiIiIiIiUmMqJkVEQkQL8EhDsGL6Apb37MWi/013HUXCyNN/nOY6gog4oGJSREREqm3V 58C8MvHXzhOIuHi2THfkbzb6zqGiDigYlLqveKiQrZuWuM6hkhAFHn1gUzC1zdPv0i3Sf91HUNERMKEikmp99Y9NZxW/xlE7t49rqOIVGr8nFuqbHPOu7eGIIlI7bT/z9/Kbhu/ZU7fgSzv2YvcrL0OU4mIiCsqJqVeKy4qpId3BQAZK c6TiNSd9v8P5NTmO86hkiVrM9Hk ICAKb 6zW3YURExAmnxaQx5ixjzEpjzBpjzH3lPN/TGPOzMabQGHO3i4wS3nwH7XvW88uLHSYRqdyfvn r2m3PfOeGICaRxmTb i1BO3b3qRPKbqe//XzQziPh6YefMnh2zHd8P22j6ygi4pCzYtIY4wGeBc4GegNXGGN6H9ZsN3A78EyI44mIBNTPW2ZUu22O3RTEJNKY7Fif4TqCNFCL5m4F4Oev1ofkfDv35PPF12tDci4RqT6XPZPHAmusteustUXAe8AFBzew1m631s4Giss7gIhIg TR/DMRqR Sd4Vm0bB/PzGLjZ9sYEOGXh9Fwkm1ikljTNMgnLstcPDl94zSx0Rqbff2za4jiIg0SkUFha4jSAMWuc8HQH6BVrwWCSfV7ZmcaYz5wBhzjjHGBOjc5R2n1jt6G2NGG2PmGGPm7Nixow6xpF457M x6XOHj5QWCQ/53twatd Wmx2kJCLB8b/zr3QdQUREQqy6xWR34EXgGmCNMeYvxpjudTx3BtD oPvtgFqvFGCtfdFaO9haO7h58 Z1jCYiEli7qNlqw6d/dGKQkogER6eNy1xHkAZqwdLtriOISAWqVUzaEt9aa68AbgKuA2YZY340xgyt5blnA92MMZ2MMdHA5cDntTyWNDI r5ef33iI/FzNnRAREWnIvnxrRR3GrolIMFV3zmSaMWacMWYOcDfwW6AZcBfwTm1ObK31AmOBr4HlwPvW2qXGmDHGmDGl521ljMkAfgc8aIzJMMYk1eZ80rDMm/ACQ9f9k9Rnex7xXP6 HAeJRALv/cXTXEcQqRFv8ZHz2fx P3t3a9i21F7KHi8JPtcpRKQ81R3m jOQBFxorT3XWvuxtdZrrZ0DvFDbk1trJ1pru1tru1hrHyt97AVr7Qult7daa9tZa5OstSmlt9UVJfiL8ip8bsHbD4QwiciR3pz/HVN WVrn46zYub7uYaRRMxGBWuagev735JH7TU56 G9sPn4Im1etD2kWCa6AraAhIvVadYvJB621f7bWlm1YZYy5FMBa 2RQkolUwp9V8T58xlsQwiQi8PTUDznq1QHsyStZZOepReO4bcrlQEmvzPDXr63VcT/75d2AZRQJhY5vPcfq2UsOeSxi2g8AbF31i4NEEgxZewvxLVVvs4hUv5i8r5zH7g9kEJGaGLrlDdcRRMq8veoFTISPBVuP/LD8yfIZ7GF rY5b5NlY12giIee95tJDH9jfhWU16a2heOnZeUSVuyh/8OnPSCS8VFpMGmPONsb8C2hrjPnnQV vAdroR8JUyTuNt7iIRU czoqZ3zjOI43Fg1P csRjuYX5dTrmt6sX8MO6JTz6Q62mp4s4Z0uLSc 94xwnkUDxFbqbwKjhtSLhJbKK57cAc4Dz4ZB17XOAO4MVSqQuhmx/nxlvtqbDiVfSr2A2WybdBsetdh1LGoG9ZgljJ/zjkMdsHS j3zV1LNZTMpzsQbSPn9Qffr fdfOXk755FQAxfi8ZK3 hXY9OjpNJfaZiUiS8VFpMWmsXAguNMW Xrr4qUi8cveb/MddbSBvXQaRR8JkDvY8/7nq57PZNnz3Jqqy6L8Sz3xlv3sLfzniAo1p1CNgxRYJhwv1PEjvlf7TdlXHI4zkXnIN38WIio0o fqyetYi09q1p2lr7Q9cvquhEpESlxaQx5n1r7a A caYgy vG0q2n wX1HQitRRtfAzdULLQsNEECwk2T/mLTM/MeqvOh97fKwmw1T dy78axZIb5lbyHSLudfnktQqf /KWe g6/WsiSqckLGuSyolzp4comYiIBFJVw1z3T3AYGewgIiJSNRNR5DqCSJ10nz7pkPtp /aU3d6yeiNxyU1IbZEW6lhST3z67gruffB41zFEpFSlC/BYazNLb 4ENllrNwAxQH9K5lOKhNzsv13mOoKIM3G rq4jiARN9nlnsvKMs1zHkKo4HOWakKHtv0TCSXW3BpkCxBpj2gKTgRuA14IVSqQyx2RPqrqRSAOV71nDK3O0QrEcsGrmQj6/ AaW9 zFhiWHLjZm6slqJfOO6l92O7kw12ESERGpieoWk8ZamwdcBPzLWjsK6B28WCKBUz8 SolU39 W3oXX525pfgkv2TffSLelMwBY sGEQ56r62rCoRLnLcJbrHX 6oPV67PwF v1R0RKVLuYNMYMBa4Cvix9rKr5liIBN /rN11HEAkLt3zxtOsIEiYSivIO3PH7D3muYG/96eWbNOpa1xGkCjPmZfLNE/NI3qnCX0RKVLeYHAfcD3xirV1qjOkMfB 8WCLlMwtrvnF7K3YEIYmIW0v2zHYdQcLRQT2Rfr f LtvdRimZrqsme86glRh8tsrXEcQkTBTrd5Fa 0USuZN7r /Drg9WKFEAs1bXERkVLTrGCIBkxexil15OaTFJ7qOImHEcqCYnPfF9zRxmEUaIgPUj6HTIhIa1eqZNMZ0N8a8aIz5xhjz3f6vYIcTCZSMx492HUEk4MZ//6rrCBJmbEEhMz8oWaSsOL/ rnqZtWO36whSHi1CICKHqe4w1w A cCDwD0HfYmEVLP89bX6vnT/psAGEQkDP x8yXUECTNdJrxN0kN3svi7maSMv9t1nFr76W8vu44gYeznOZlVNxKRkKjuIjpea 3zQU0iUoW83GzaW21vKiJSlb1bttHUdQiRIJny2nKGDm7tOoaIUP2eyS MMbcaY1obY5ru/wpqMpHDrHjhmjp9vz1slUMRkYaqYE W6wgiQRPhD/68zf 8upB//kMLnYlUpbrF5HWUDGudDswt/ZoTrFAi5UnLW1en79 1VUNdRaRxaPPs464j1EnnjzUfWCoW7zf4g3yBuGDmLjzLc4J6DpGGoFrFpLW2UzlfnYMdTuRg8f667ZdmtQKdiIhIg/Dlt7 4jiAiVH8113hjzIPGmBdL73czxowMbrTAKij28e2yba5jSB00Z4/rCCJh53dfPes6gog0FmG0mmvmJvUaioSD6i7A8yolQ1uPL72fQckKrxOCESoY/jRhGe/M3Mint53AgPYpruNINeTvy2HVrElYn5eC1T8ypI7H 2Xm56zN20Pb4y6ifde AckoMnfzWqfn/3b7C8BtTjOIBMOM9ycSl5pMXFIC3Y/r7zqOhJmc3fV36xuRhqS6xWQXa 1lxpgrAKy1 caYMLo VbWMPfkAZOUVOU4i1RX3dDsC fHh2IUPAbB99WswXsNjJDBunjS2 q kQfLpshlc2Luul1tEwkvyH 8CwAewYrnTLFIiJSd8FrKLXZfnOoKIUP0FeIqMMXFQMunMGNMFKAxaqiDYX/laTZurF/Zm7QraseNtftCOLY2PD/dXx6dtXOA6gjiycZnbnnEREWncqltMjgcmAe2NMW8Dk4F7gxUqGCJKq0ktwlI/rHx1TNCOnWDyWbNwWtCOL42N 9eU6Vt/dB1BHFn8u/tcRwiJH1/ L/O /MF1DGnEps7YjM8bPj2zIuGiuqu5fgNcBFwPvAsMttZ H8RcARdROipXWw3WD8dkfxPU4/u uCuoxxcJpRyzhNkZa1zHEBcayXCbFs MJ 6u37iOIWHm5zmZITnPDz9tYtFrK3nh3/NDcj6R qS6q7lOttbustZ aa2dYK3daYyZHOxwgbR/iqe/kbzxSuV6eFfA GQWfvee6yhSz/kjgzckuyZunDzKdQSRoJtwz2NM/tcbrmM0SnMWbnUd4QjzXg7 XNonHp7G0gXbS 4szg76 QJtR04hK7buDdrxJy3JZOOu4M1ftSH63L5yaw7FPvU41Ualy0YYY2KBeKCZMSaVA1MPk4A2dT25MSwbIFcAACAASURBVOYs4B AB3jZWvvEYc b0ufPAfKA662182p3rpL/ lVLykH6T7mFDW17ktKsNcu/eo7jrnyYWR88RUzTDrToNoiC3CwiPB4yp75Ox8xJbGx2EkN2fMDPba8nfvdy t8b3B7Uhmj1zkw6pbYg0uMJ6nlW7MjA6/NzVKsOdTrOvsJCTATER8UEKFnwvL94Gk3jEvn7rFcZO/haBrbpQsuEZNexqu2nDcvplNoSn9/SNimVIp X2Kjosuf9fj9 a4n0eMgrLuTOr/7FM2feRmJMXMAybP0lg1ad2lX4fM6evTRJTiAiorqzRIKr5ZbGNWeyyxdvAZB9xXkkN0ulML AmLhYNq9aT4uObfBERZK9fQ prdICds75X01hz9r1nDr22rLH1i1YQUFOLnHJaRTsK2TW6z9x/p8uIjElkS//9gXJbVI4 ZpTan3OBUt3MOOnDMaMHhiIHyEgZj6/zHWEcj1962Ti/QY/lrxIw8CLuxAXF0m3zins2lNAsdfPgD7NAfjr0zMp3LSP /95arWPn7itCLYdWLzxvY9WcPnFPcttO3XGZha9tpKhtx3Ftm376HdUc9q2Sjii3XdTNxIbG8nSxTvoP6Ala9bs4ZILu/HyfxZx6WW9aN70yNe01euzsH5L1/SS1/Tnb/2B7LRIrh07EK/Xsn5jNi2ax7Moax b9uTxxcJMNu4 UOR9d9dJZOUX069tMvM2ZrE9p4Cx7xzZ03rFse3ZtreQV64/5pDHs/OKiY6M4IO5mzi5ewtuf28 I/u15tEvSwr6afeeQovEWKIjD31t3F gRXkiyC30ctTDX/P8VUfzm7fn8X X9ufiQYe 3i7KyCLCGNo3jWfCoi088MkSHjy3FzcN68yKrXs56 9T XDMULq1TCQ5LuqI/JOWbCUpLpIXflzHK9cNpthniYs 9POGtZZT/ 9HxpzUmWM7pfHh3E08 /1azunbiomLt3JazxbkFfl44epBFPn8RHsiWLF1L8d1TiO30EtCTEn5tKn09xsRYYj2RJBf5MPjMbRNKfn/t3ZHLpt257Fyaw4vT/uFZy7tj9fn59evz G6oR05rVdLerRKpGVSLAA v8WUHg9g pqdLMzIZnNWHmcf1ZoTujYr xnyirz4LWVZAC549iey84oYe2o3urVI4NpXZh3x wkGU1nFb4wZB9xBSeG4mQPF5F7gJWvt/6v1iY3xAKuAMyjZamQ2cIW1dtlBbc4BfktJMXkc8A9r7XFVHXvw4MF2zpw5hzyWft XADx64VFcPaRj2eNLt2Tz0pR1PHFxP2KjgvvhVqpn59aNNHuh/m3dsTS6Pz0KF7M2qju5R4/GeotIatcbYwztew4iJiaOBc cS7urn2fnppVkr55OYqdj2bt2BkOv/fMRx8vJ3k3Wjs1Yv4 87J30POZ0dmxZz/pZX9L1hItIbd6ajasWkNqyA1HRMcTGNQFKtlQpKiokObXkRWf3viK XLSFt2du5MmL 1Hk85MQE8mevCI27Mqjb9tkMrMLGNA hbwiL09OWsGfLziK2CgP wq9tEiK5eHPlrAnr5iHRvZm5i 7yt6ATuiaRu/WSfz rJ5EeSJYsjmbBZuyiI3y8Mn8DH53Rg9aJ8dy/BPfEdv2LaKSlpT9fMaXgrFR CN3MDDxMm4bfBlJsXE0jU/ktA H8 rpH7E1Zzd a4mNiubuqWMYlHo zeObMWvrdIptETlmCVemP8hZ3Y6la1orRn/ OA8Mu4lrvhyNN3IzAG08w3j/kmeY8ssSiv0 UuMSGD/1GWI88WT6pjG6 6MMbd Hpds3cE73Y4iMiCC7II tuXv4YNl3fLPtOQDeO2siiTGxrNyxmV4tOjB1/WKemvMXvJFbgvp3FSjp0aczfvhY3ln8Dd9se47 CZdw08CLueP7cfgpwkM8yZGt6d20P1N2vknLyEEU vaRbRbj8bbCF7mVazo9xK78bCZu/SfDmt7I1N2v4PG2wBe5/ZBznd/mTlbsXsWqgi/BFweefDpGncb6wimYiGLwJRBpk ibOoz5Of tNLe1BmMsl7T/PR9uegqA1p4TyfSVzHs23qb0SDyBDbnLyfes4ZSFfq6aHEfLN99hyVsf0uXzNw853ur w i2cOoRj3VY/DMxfm FOfbEJbM3tQUdt6wue2xdxz503rC03Pbr0o/CF9 EbstmsqtJKmn79pAx m6Kdu6k88evkRsdT9F9D9P0T/fyS7ueeHv2odv/PmLVsafTYe4UYn0lH143p7Wj7a4M1l54LV0 fYPM1Nak5OxmT1IaHq Xlnu3l3v xmpDm2503LKaVSedT/cfPyfnsb TOfFrEvr3I/mlf7Bl2Nl0  4TVvc6FtO5K12/fKfCY2WOvZ ivTl0fKPko86G1l3pmHlgOHlOQjtmD76/0jytWu9ka2Yz4ot odC0xOPLoyi22RHtInyFeLx5FMekHvJ476u7snLZLnYvzwKg8wmt2bhqD0nNYvHN20NWnCEl/8DnuLz0ODwb8oixJR/XBo3uxZT/LKOJr R VhNDk31 ojDs81ia AzFxlLYJpaEzQfWVfQcnYpv3oG9nYu6JxC9KrfSn7W iRiYQkxsJPk/76zx92anekje4yO3bcwhv7dAKeqegAESU2PYu6cQX5GfuPXV6/2bFVPMj3EVv5bV1JDOTZmxbnfAjleZSXcM46y/T626YakLBrThswVbiPZEUOTzM6xbM6auPvL/Z5vkWLZkFzC0cxo/r6v9aKLoyAiKHM bvf20bvxz8oH3oZN7NOeHlTtIjIkkp/DI/ 8bnhw511o7OJiZKi0myxoZ81tr7b8CemJjhgLjrbVnlt6/H8Ba /hBbf4N/GCtfbf0/krgZGttpYPkDy8md YWMvjR/wFw/fHpjD /D3CgwCzPqT1bMP68PkRElAyRnbx8G t27KNPmySshSmrdzCgfQojereiyOfngzmbeHvmRh4a2Yt7P1pcq98JQO/WSSzL3EtibCSFxX76tktm7oYDL gtk2IY1DGViYvLH26SQg53J/6Pi4s Y2VsfwYUzgZgs2lJqj LeFPIGk8XuvpKrmYffHtO0hkM3vttrbNL LmocDzzbHfXMTBRu0jo rTrGNIIvP944D5EiVQkPzaNn4f8yXUMkXI9naJV66VEKIrJau2OZq39lzHmeCD94O x1tZl4kJbYNNB9zMo6X2sqk1boEYzro9//Luy20Wl3e2VFZIA363YzncrKr/iO2FRZln3/n51KSQBlmWWjGvPKSj5QHRwIQmwbW9hhYUkwILYW6AYMJQVkgBt7bayfuX9xePht1VINjwfx4wnvaDiK  h4okJv7k2IiK1pUJSwlnH4gg2RGn n4RGtYpJY8ybQBdgAaX7B1OyHn5diklTzmOHd5NWp01JQ2NGA6MBOnQ4MEfKWstD5/XmoU9Lhtdl5xcfUkhOu/cU2qXGA7Bmey6n//XAEvv3nNmD5gkx7M4r4omvVpT7Qzx8Xm/SEmK4/d2SYX//d2l/npy0giuP68DNwzoT5YkgylM6vCSvmCKfn1d  oVP529m297AD4140XsuoyMrL5RFQs1fdOSwLhEREQm8s/OieSHZ/R7I0jhUd5jrcqC3DeCSSqEc5grl90R NW4YvVon1fEnkbry 3xEHLQYS1FhAdGPt3SYKLDmJp5KhK AgXnTy33eayNYdvJLFOzOwLc3E/w hm56iRmdb8fEJODPysAkNKf14PPJnPwc0fsy6XXrO8Q93Y7ZAx/nmPn3k0lzWrODLdfPYtPMT4iITWK3pxnjpxWQyYGFKE7p0ZzuLRP5z7Rf8Ppt2Vj7w3Vu1oSICMOa7bn87bL TFiYyeQqeur3axLtYV R74jHjSeHhO6Plf9NvkTw5BzyUJyvG/meA/MCjLcpNrL28zb2z/EDSLRHkWOWVNg22fYj2ywq97kWEUNIiU5jZf5XGKMrv Hmzk98DF1x6FvVquEj6T5lgqNE0hDNHfg7spO7ODl3ThQkFjs5tdQTPix/TVExKeE1Z/ID4PaqirgandiYSEoW4DmNksV9ZgNXWmuXHtTmXGAsBxbg ae19tiqjl2dYvLgHkkJQ NDuwLljJaXE5ObgTcqkWOyvgJg7cVfkz/hXvJ7XIRv61I8bfrRpGUXivOz6T/lFgBWX/AFqa070bR5m0MKYr/Px4YVc jU58DIbW9xET6fl5jY0P/deX1 Ij11W32y0OvD57cUey3J8UeuoFaZrdkFzN2wh9N6pRHtieSlOZNok9gcsGzI3satx55btjpmQXERkREeIj0evlw5h915e7lm4IGV97y kkK1otVgl27bRK/mbRn9xdPcPPAierdsF9DVPvdbtHU9V319XsCPGwjHJl/NzUdfyOM/vcgVfS7gjK4D c cr/jdCReX/d725OWSGBNXdr/I6yU6smSwyoeLf KULv1Ijokn0uNhxJtjOCP9dO4ZdgkFxUXszMslKz Xns3b4vX7DllxNSt/H7lFhbRLbgrA/C2/4PP7mJ 5lpuPORMo X 8eNtG3lo0iSZRcXyx5e c1WosrZo0Y0febm479iIWb1vPOT0GlR130db15BYWMKBNJwCmrl/G2t2bSY1N5Ir Jx3xO/D7/YesuFqQl4 v2MfMtz4hd8kyEvv3pf0x/WjesS3T/voinoREEju04bjLRjLtlQ8Y9utf4Yks d1sWrGO1l06sHNTJvEpSSQ1TWbrLxmsn72IwaNGsDNjK606tcNb7MXv8xEdW7Ly77b1W2iZXrLwefbOPfj9fhKSE4nweCguKiIyKorIqPIHCG1dt4k5z7/BWX 5lwhPBPuyc4mKicIYw8Yla1g3eQptjxuE5zfX1/Cvo2HYntSck2ZNwe/3Y/0Way3e4mKytu2iRcc2zPzgK4679Gy2b9hyxOq8G5auIWvLNvqfcQKrZy2iRef2JDdLZcemTDYvW0vmU0/T9v576XniIFbPWsSuNRsYfuMleIu9ZK7dyPwHH8M2b4ntcDTZ67PJNZ3oe5KH4VecxLNjSqbV3PbCqfh8PnJ357Ly5xWkD zMqukraNurHZ36dWLK2z9SmFtAlyHdiEuMIzI6isTUJvj9lo1LN9HtmK74fJbo6MoHkK3dkMWHby7Dm1vMqZd0p7DQR98 aSxcvIO4uCjatUkgNjaS5k3j2Lknn1f/vYB77hta9v1ZewtZ 0sWefleoqIj6NQ ifi4KAqLfbz4yAySCyzpF6Wz8Mv1JAd MFXAdb2sM1sycsj7aQdZSR48 T6aD2pG564pTP90HSm5JRcA93WI4/d/GEpefjHFXj/JiTE8dt P9D25HZvWZuH1 olcfuACZ1aKh5SsAxdJc1rHcN/DJwCwZMUufvz7Qs59YDBfPnbgs6evdxIjzunMNxPXEdckkuwdBRUuouMZmEpy01iuuLQXxV4/23bs47OPV3HVNUcxZXoGG1bvoXf/Fqx4Zw1FxtLjok7Exkay/O0DC0J1vqQT6z78BYCmF3fEkxrN7z8suSj6/i1DaZUUy/WvzWLT7jzeGz2URRlZPPJF Sv0fj72BHIKvAzqmIrfWlZvy WjeRlcd3w6n8zbzP/7/sh9jbu3TODzsSeybW8ByXFRWAvHPT6ZIq fwR1TWbIlm0sHtee2U7ry9swNDE5vStuUOBJjI4kwhmYJ0RhjyC308uKPaxnYIZUt2fkMaJ/Cuf8sWXDt6A4p3HpyV256Yw4PntuLR79czts3HUfzxBhaJsWSV Rl295CLnz2p7Jcn9x6PP3bpZStkOor3dbhqUkr PeUdZzbtzVfLs7kqLZJLNm8l3duPo4rX5pZ9v0f33o8s3/ZzcqtOXw8fzOjBrblk/mbufXkLvy0dhePnN H2KgIYiI9JMVGkrEnn0iP4ZHPlzFr/e6yFWJbJsUcMiJxxZ/PYvW2XMa8NZftOQXMeeAMoiMj MvE5RgDW7IKKCj2cWrPFpzRuyVtUuLwlP4M1lo63T Rh8/rTduUOOZu2IPXb7lkUDt 89Zc rVLYW9BMbed0pVjO6WFTTH5PTAAmAWU/SastefX6eQlq7X nZKtQV6x1j5mjBlTeuwXSrcG X/AWZRsDXKDtXZOhQcsVVUxuf6Jc sSW0IhhMVk/j0ZxDVJLLufl5tNZFQM0TGxIcsg9VNBcRHHvDOo6oYh0D7yZCZeFdB10qQe OyasXSfXa 2fa6TzNQ2 C65gjPuusl1lDKFeYXExJdcQMjekU1cUjzRMTW74FYf/P2vs4gKwxVde17ZleHHtyMqsuoLpnn5xeQXeklLqfoC4/4LA/np8Vx6dW8mPFryufKcPwyiU4fyP6OsWreHxUt2cPH55S985/P6MRGwduNekppEsWNXAT6/n/69m1eZB CTCavp0b0pvbunlf08Ho8hpoqLDlXJKSgmJtLDe7M30jwhhrP7tq7R96/cmkPLpBhS4qMPeTyvyFurC9CB4PdbtuUU0Dq55heTf9m5j2YJ0STG1j53odfHvkIfTZuU/E6stWV73oeSMSY8FuABxgfj5NbaicDEwx574aDbFrgtGOeW8JZFAikE903L/nEPJiKCw19m4uvRvnzi1sE9cq69OeovriOIA20vGQWNqJg89efw 1n3F5IAyc0b7vvHccPbM2/V8qobhthpw6u/l3B8XBTx5exNWJkxvz2ahCYHXusrKiQBundOpXvn1Aqf95QWvN3SUwBo2bxJjbKMGtntkPs1/Vkqsr9ounZoeq2 v0erxHIfj4 OBEdvkxERplaFJECnZjX7/1KemEgPMZEHRlC5KCRDpbqruf5YdavwZgwEbsanNAQmTDYdF6mrrjFnkxZf/pu5NGwRkY1nf SET7W4nEvt2iQwz3UIB/YXkjHHpdGjlxaTEzlcpZ mjTE5xpi95XzlGGP2hipkIEQ04CsCDdGqvncDkDV2Jb4Hd/Fz 5sdJxIp33EpV7uOwCeXP U6gjhiGngxGfvfT8tut /Z2WESad8m/C5YDflNn6Adu8kJzclpdaBb7aYb jNsSNugnU kvqq0mLTWJlprk8r5SrTW1qtlUMecVPImlBhbt3HlEhrHXnwnjM8mpVkrPJGRDP31MwE9/qLYoA4fl0ZkaNsBriNII2bK3UGr4ejUvwcd58yl4 wql0uQRmhQ/ Ct/H79NX25b/yJQTu SEPRaMb5DetWMrm5RWJMFS2loZvR4ld0  2nVTcUEQlzJqJhF5MA8QnxxCfWfQ6TNCxZCY3mI6xIWGs0/xL3v91G1XF7BHFnfUT1J9lXpuuFDx6yeqtIXTT0niEJbw15UYcN1/7WdQQJY797VL2GIuGg0VRW /eXUTFZf6WO/S4gx2nWpmNAjiMC7j/M39EnsEPApX7pNazhDtmPb93CdQQJY3GatiQSFhpdZRXpabhXcRu65KbN2UvdhjrtJT5AaURKNImu3dLjgRLtaXj72Un1RcWEz/Y0gXbidRe5jiBhqtPF6a4jiEipRlNMen0l 4JEaTuIeq2A2s95nZF GwU3Tw9gGhG4pM/xTs fFKO5ZNIwRej9Ouzkp4fHBdmkJK1/IRIuGs0rtdfvByAqUj2T9VnhVbVfOGfI9X hRdtOAUwj4v4Db/dmWqpeREJjwNDWriMA4NGUJZGw0Wj NZb1TOoFqF5r362/6wgiYcP6Y jVop3rGCKB98rbrhNIOax1naCE0Uc5kbDRaP45ev0lr4CeBrzynVRsD1q9VRqeo5MvdB1BJCh6HX 06whSDn YVJPZWYWuI4hIqUazFNb F0DXKy KG5F3LnIdQSTgXrngftcRRAIq5y//YNv0mfRyHUTKV1pL7o2GpCJ3MfRZTiR8NJpi0pYWk41gf2cpR2JyU9cRRAIu0uNxHUHC2KpTL6T7d7WfZx5KyV98TZtupXsJXzTCbRipWOlnKBthKKssHRjYX9vGiISLRjPMtXSUKxG6mlXvLYkZ4DqCiEjYS jZg3Ud 7iOUaU151xxoJCUsHbqsA4UdIpnxA3u o73No ieVO3WzKJyAGNqJgs7ZlsND xiDRo/ljXCSRM7I5PqfC5yFNPD2GS2ml6dD/XEaSaoiIjuOveIXTtlOosw9h7j3N2bhE5UqMprfbPGdc4 /ov7coXXUcQce7VM951HUHCxKBp35VbUDbt1jms3/N2/P7PJHz6JSdcrYWk6pukhChn505MiHZ2bhE5UqMpJst6JsP4jVWqp3XHHq4jiByiib n6wjSiMXGx7Gz36G9NZFvfsDAs4dDmL7nrTrhbIbfeAnte3Z2HUVqISIigqxkzdkWkUZUTPZqnQTA6b00aVtEAuvirr8K TkTYjTMVQ44 4UnD7nf7ZijgPAcjZP9p//jnOeecB1D6ir8/rRExIFGs5pr95aJrPjzWcRG6UqaiASWDfHeazd3e5SezduF9JwS3qJjY8p/IsyKyfVtu3P2r85xHUNERAKk0fRMAiokRSQobjj6LOL93UN2vtuPvyBk55L6xWciKP5//3Edo0IDXvyX6wgSIGF2nUJEHGk0PZMiIsHSPCGJmTd8RN/X 7qOIo1Yp3nzMMYQExeeQ6A3j/k9vbpoC5CGwmqcq4igYlJEpF4JZQ o1C x8eXsvRdG3Uen33GD6wgSQC7 tJqPaBv6k4pIpRrVMFcRkWDqETcy6OeYecNHQT HNCBhUkyuOkHzJKXuTjhexaRIuFExKfXanMTTjnxwfDYL4oeGPoyISJgJh9VcN7VIZ9ij97mOIYHm4BOktncTCT8qJqVempVScpXbG5dW7vOayyEi4t7qo45nxJSvaNq6uesoEmBX/2aA6wgiEgZUTEr9FFH yrxzE04uuWH0py2hZ3QRQ8KN4z/JpucHf i3uNGhbVLIz kP8TZMIlI1J5 4jTFNjTHfGmNWl/43tYJ2rxhjthtjloQ6o4S3/T2PtsmhV7vTzn2o9HkVk9LwpHK06wgiNZLSvo3rCNKAaJirSPhx9Yn7PmCytbYbMLn0fnleA84KVSipPwwlVycj4ptW1KDMiqjeIUgkEnwWv sIUs 4nDPpe/41jjrlOGfnl4ZHtaRI HFVTF4AvF56 3XgwvIaWWunALtDFUoajqQz7i27XRwR4zCJNCaW4A7B6tNUc5SkhhwO VchKYGmYlIk/Lh6l2lprc0EKP1vi7oe0Bgz2hgzxxgzZ8eOHXUOKOGtogV2UlulA9C1/4ksPvUN4EAvpkh999y5d7qOIFIt6y7SnpISeOGwOrGIHCoyWAc2xvwPaFXOUw8E43zW2heBFwEGDx6s6qGhq Bqe3JqswNNSt9zjCbsS4gEcwGeZ45/g4gIzQWWmgn1Z /iCA /nHwe5z16d2hPLE54jk7FN2 P6xgi4lDQPplYa0 31h5VztdnwDZjTGuA0v9uD1YOaZh6XvkUM5ueT/ Rv6m4zZCzmZU6ktQrXwphMmnMTuoQ2P1Nrf/A9b4eadqsW2ouNq3c9e2CYsM1t9Fj7hwueO5xXfhoJNqnJ4f0fDEx5a/kLiLuuHq1/xy4rvT2dcBnjnJIPZWc1pLjbn T2LgmFbaJjIrm2HFv0ya9RwiTSWM2dsh5AT2esXFE zoAEOkJ2kASacCGXHZuyM4V37oVMXGxITufuBfKnu/8CEtyotZAEAk3rorJJ4AzjDGrgTNK72OMaWOMmbi/kTHmXeBnoIcxJsMY82snaSWsZZEAwF4qLixF6qMYmvHe S9zTaeHaJdcwcrFIpUIZQ/hiddfFLJzSeNTmKgLaiLhyMm/TGvtLuC0ch7fApxz0P0rQplL6qf8G75j1SfjaT9qPKHfQlnkML5Y8BQE5FD/veBFuqS14vfDfxWQ44kEk4a2iog0Pnrll3qvdcceHHvHu7TuqOGs4t6H5wVu1H6XtPLWMBMRCQ9aXVVEVEyKiARQiybqH5fGpzBCQxAbI9WSIqJiUkQkgBJj4oj0tqJD1Cl1Os6pzW8JUCIRkeDo2qVkteAmJzR3nEREXFExKSISQJEeD/N//S13HHNjnY4T7YkKUCIRkeDomp7Czf8czrVX9XEdRUQcUTEpIhIEZ3QbwJ PeYkecSNr9f3W2gAnEgke/9P/ch1BHImOjtTiSyKNmP71i4gEyYW9h Ax2mRbGr6jzz3ZdQRxLGFYC9cRRMQBFZMiIkF0We/QbRovIuJK126pQTlu25EdgnJcEQkMFZMiIkF0UZ hvH3mF65jiIgEVbC2CYnwaMlYkXCmYlJEJMj6tUp3HUFEJKgG9W9BdrMosmNcJxGRUFIxKSISCr4E1wlEgmJHYjPXESQMxERH8odHhxHVKi6wB1bHpEhYUzEpIhKGLFrNVcJfZmprhs e6jqGhJHRY48O6PGSEku6OqMStV2SSDiKdB1AREQOiPZ1pFl0Bx4cfo3rKCIiNZacGNhxrief0I4dO/Zx9hmdAnpcEQkMFZMiImEkNiKJr69 znUMkWrx3PQb1xGkEbj0wh6uI4hIBVRMioiEFQ1vlfqh3c8z6ZWa5DqGhKF253Ug44uNrmOISAhozqSISAjE0bKaLbXahNQPQdoJQhqAxKRo1xFEJERUTIqIhMD7F75YzZbqmZTwsi8q9ojHNlxzGwkp6pUUEWnsVEyKiIRAetMWriOI1Io1R35UOOuBsQ6SSH2hXmuRxkPFpIiISAOW98xzbLj2t7X /sP7ytd2GVC3QNLgdU5PCchxzv/jMQE5jogEj4pJERGRBmzQyFM46w 31v4AB3Uz7Uhsxsgv3w1AKmnIOrYrfwh0wrDqj9DIi7C0b5MYqEgiEiQqJkVERKRa8hJTXUeQeiq3TQzXXXVUuc8VdUs44jGNlBWpH1RMioiISIUKDlqAx2oynFRTbtuYQ 7f/YehAEQMOHII7J13HUtehBYfE6mPVEyKiIRIjK j6wgiNdbhlVdcR5B66J4HhrKv/YEL41kBwQAAG7hJREFUEZ7Iko crdpp6KpIQ6JiUkQkRObcOMF1BJEaWXXqhaT36152P2X0aIdppD6JiIjg1juPoelprbnu/06s hsO6/RuOaxVcIKJSEBFug4gIiIHDGk13HUEkTIXPPf4IfeHXnauoyRSHyXER3HFpb1q9b3XXNEnwGlEJBjUMykiEkaeHnGL6wgiIkFjKlhaRzMmReonJ8WkMaapMeZbY8zq0v8esTycMaa9MeZ7Y8xyY8xSY8w4F1lFREIpIkLX CQ87IlLdh1BGhF/89iqG4lI2HH1qeU YLK1thswufT 4bzAXdbaXsAQ4DZjTO8QZhQREWm0si 60nUEaUTG3Xec6wgiUguuiskLgNdLb78OXHh4A2ttprV2XuntHGA50DZkCUVERBqxqCStuimBV9HuMnGxWsZDpD5yVUy2tNZmQknRCLSorLExJh0YCMwMejIRERHhlNuudR1BRETCXNAuAxlj/geUt67zAzU8TgLwEXCHtXZvJe1GA6MBOnToUJNTiIiIyGE8kR7XEUREJMwFrZi01p5e0XPGmG3GmNbW2kxjTGtgewXtoigpJN 29v 3d dRUlR3/8ff3xlmAcaRYVM2WSYsilECBCFgcIBIJEYwylFRA/FJSGL0/MAoSoyCqAg Lgl4MCFG4oPLIxAf8HBElGFxQhBBQAQxggENgmwDKIIsM9/fH13Tzi6DdFfPzOd1Tp2uvnWr lZ/a6b79r11r7/0Na83HZgO0L17dw0KJiIiIpJoNMaYSI0S1p/0y8DwYH04MK90BjMz4K/AJnd/LI5lExGJmbu6TA27CCIioUlLLdnifeYl5XViE5HqIqzK5CTgB2a2GfhB8Bwza25mrwR5egM3Av3MbF2wDAqnuCIip0erzCZhF0FEJDSX5rQu8XzoTzqEVBIROR1CGTrL3fcB/ctJ3wEMCtb/ARXMbCsiUk1VNGE3gJ0oM WuiEiNUnou3cLCkAoiIqeFeq6LiMRRUlLFlckb2t8Sx5KInLyDaRnszGoedjFERCTBaFIfEZE4qqxlMiU5JY4lETl5Pd9ZFXYRREQkAallUkQkjpIqmLG7W a13NrzijiXRkQkXEn6JipSrallUkQkjqyC3/D dmWVpuAVERERCZ1 DxIRiaPkSu6ZFBEREalOVJkUEYmjrs3blUlzVwVTREREqh9VJkVE4igpKQkvLDnQTnphm3AKI7VK69Vvn3TeY0m6C0Zi70CDZFLq6KuoSHWmTwsREZFaoF5GvZPO2 j/Xo5hSUQi7p7UN wiiMg3pJ DRETi7Ft1 4ddBJFKZZ3dJOwiiIhINaDKpIhInM0ZOqlUiodSDpGKuBeGXQQREakGVJkUEYmzOsnJYRdBRCRhneh0RthFEJGTpMqkiEjIMus0DbsIIiWY6euBhKdOqq4/kepCA/CIiITop 3u5affuTTsYoiIxM2hZMgoqCSDev6LVBuqTIqIhOiOi4eGXQSRMnTPpMTSTff3In//kbCLISKngSqTIiIiIhI3TRrWpUnDuhVud7VMilQbqkyKiITgvHqDuaDJeWEXQ0QkYRxsWIcz80/gqk2KVBuqTIqIhODFoQ EXQSphQ6l1CXjuLoXSmKq16Qu5H8edjFEpAo0XJaIiEgtcezOe08qX92M jEuiUhZzVpHpgRpdk5myCURkZOllkkRERGJ6vjeRpKS9FuzxN81P nEmvYN6fptTZckUl3o00JERESiqlqRbNOmDW3atIlNYaRG2rZtG2bGiBEjymyrSkXy PHjjBs3jvbt25OWloaZMXfu3NNYUhH5OqpMioiIiEi18 ijjzJhwgSaN2/O7bffzrhx4 jUqRPjx4/HzFi6dGnYRRSp8dTNVUREpJbwQo2SKTXH/PnzycjI4PXXXyc1NTXs4ojUSmqZFBEREZFqZ8eOHTRq1EgVSZEQqTIp1d7777 PmdGvX78K83z7298mJSWFTz/9NJq2cOFCBg0aROPGjUlLSyM7O5s77riDAwcOlNl//fr1XHfddbRp04a0tDSaNGlC165dGTVqFMePH4/JeUntVPxeom3btnHttdfSuHFj0tPT6d69O/Pnzy zz9GjR5k0aRIXXHAB9erVIzMzk4svvphZs2aFcAZSE7k7TzzxBJ07dyY9PZ0WLVpwyy23cPDgwXLzn w1eejQIVJTU ndu3eJ9CNHjpCeno6ZMXPmzBLbpk2bhpnx9NNPR9MuueQSzIwTJ04wceLE6D10rVq14s477 TYsWOn6Z2QWDp8 DAPPfQQXbp0oX79 mRkZNCrVy9eeOGFEvlGjBiBmbF161Y  ugjzAwzi96/e9999wGQk5MT3WZmYZySSI2nbq5S7XXq1ImcnByWLFnCBx98QIcOHUps/ c//8mGDRu46qqrOPvsswGYMGEC48aNo2HDhlx  eU0bdqU9evX88gjj/DKK6 wYsUKMjMjQ5OvX7 eiy66CDPjiiuuoG3btnz22Wds2bKFadOm8cADD5CSkhL385aa7aOPPqJHjx60a9eOG2 8kfz8fF588UUGDx7MokWLyMnJAeDYsWMMHDiQZcuW0alTJ37zm99w PBh5syZwzXXXMO6deuYOHFiyGcj1d2oUaOYMmUKzZo1Y TIkaSkpDBv3jxWrlzJsWPHSrQMVeWazMjIoEePHqxcuZLPP/ cM86ITA2xfPlyjh49CkBubi433nhj9PiLFy8GoH///mXKOWzYMPLy8rjsssvIzMzklVde4eGHH2b37t3MmDEjZu PfHMHDhygX79 rF27lq5du3LTTTdRWFjIwoULGTZsGBs3buSBByLz8w4ZMoQ2bdrwhz/8AYhcnwANGjQAYO7cuSxbtozhw4drcCiRWHP3Grd069bNpXaZPXu2A/7b3/62zLbhw4c74K 99pq7uy9evNgB79Wrl /fv79E3hkzZjjgo0aNiqbddtttDvjcuXPLHDs/P98LCgpO89lIbbZ161YHHPDx48eX2Pbqq6864Jdddlk0beLEidG048ePR9N37drlrVu3dsCXL18et/JLYnt/xVp/r2OnSpfSli9f7oBnZ2f7vn37oulHjhzxnj17OuCtW7eOplf1mrznnnsc8Pnz50fT7rrrLk9OTvacnBxv2bJlNL2goMAbNWrk7dq1K1HGvn37OuBdu3YtUcZDhw55dna2JyUl c6dO0/tTZPTruj/3PDhw6NpRZ/VkydPLpH3yJEjPnDgQDczX7t2bYltrVu3LnHtFRk3bpwDvmTJkhiUXqT6AFZ7jOtdoXRzNbOGZva6mW0OHrPKyZNuZm Z2TtmttHM7gujrFI9DBkyhObNm/O3v/0t ms2RH7pnDVrFtnZ2QwYMACAKVOmAPCXv/wl itmkREjRtClSxeee 65Mq9Rt27dMmlZWVmaj01ionXr1vz 978vkTZw4EDOOecc3nrrrWja008/jZnx2GOPUafOV51NmjZtyj333APAU089FZ9CS8Lr2LNLlfcpatG7  67adiwYTQ9PT2dhx56qEz ql6TRS2Mubm50bTc3Fy6devGVVddxfbt2/nggw8AWLduHfv27Su3VRJg8uTJJcpYv359rr/ egoLC1m9enWVz13iY9  fTz77LN0796dMWPGlNiWnp7O5MmTcXeef/75kEooIhUJq5vrXUCuu08ys7uC53eWynMU6Ofuh8wsBfiHmS1w9zfjXVhJfHXq1OHnP/85EyZM4O9//zvDhg0DYObMmRw5coSRI0dG75dYsWIFKSkpzJ49m9mzZ5c51rFjx9izZw/79u2jUaNGXHPNNfzxj39kyJAhXH311QwYMIDevXuTnZ0d13OU2qVLly4kJyeXSW/VqhUrVqwA4PPPP2fLli20aNGCTp06lclbdB/x2rVrY1tYqdY 6PJ9kr48wrfeX1Xu9jVr1gDQt2/fMtsuvvjiEhXGU7kme/XqRd26daOVyYMHD7JmzRrGjBkTzZ bm0uHDh2iXVwruke e/fuZdJatWoFwP79 8vdR8K3atUqCgoKMDPGjx9fZnvR2ASbNm2Kc8lE5OuEVZkcDFwSrD8DLKVUZTJomj0UPE0JFo1pLhUaOXIkEydO5M9//nO0Mjl9 nRSU1P52c9 Fs23b98 Tpw4Eb1BvyKHDh2iUaNG9OjRg7y8PB588EHmzJkTHQyiY8eOjBs3juuuuy52JyW1VulW8yJ16tShsLAQIDr4SbNmzcrNW5Re3qBSIlEpqXiTJlBBZbLoOjvrrLPKbEtOTqZRo0Zl8lblmkxNTaVPnz4sWrSI3bt3s2LFCgoKCujfvz/nnnsuzZs3Jzc3l1// tfk5uZWOuBaeX83RZXdgoKCcveR8O3btw IVCpXrSr/OoTI57KIJJaw ued5e47AYLHpuVlMrNkM1sH7AZed/eVcSyjVDMtWrTgxz/ MW 88QabNm2KDrxz5ZVX0qRJk2i M888k6ysrK/tA966devoPr169WL /Pns37 f5cuXc88997Br1y6GDRvGokWLwjhdEc4880yAEqMUF7dz584S UTK0 raqyrdXnT97Nq1q8y2goKCaEWgeN6qXpP9 vXD3Vm8eDG5ubmkpaVFR3gtGmDt6NGj5OXl0blzZ5o2Lfdrg1RTRdfD6NGjK/1cXrJkScglFZHSYlaZNLNFZrahnGXwyR7D3QvcvQvQEuhhZudX8nojzWy1ma3es2fP6TgFqYZuvvlmINIiOX36dAB  ctflsjTs2dP9u/fz8aNG6t8/LS0NL73ve8xYcKE6L2X8 bN 4alFjk1Z5xxBtnZ2XzyySds3ry5zPaiL15du3aNd9GkGun6o0sq3x5cP8uWLSuzLS8vjxMnTkSfn o1Wfy ycWLF9O7d2/S09Oj2/Lz83nyySf54osvKrxfUqqvHj16kJSURF5e3mk5XtEtAmqNFom9mFUm3X2Au59fzjIP2GVmzQCCx91fc6wDRLrC/rCSPNPdvbu7dy/eCiW1S// /enQoQPPPPMMs2bNokOHDtEpFIqMHj0agF/84hfs2LGjzDG  OIL3nzzq1tz8/Lyyp1LrehX nr16p3OUxCpkptuugl354477ijxxWnv3r3cf//90TwilUlr0RKAT341psy2ESNGAPDggw Sn58fTf/yyy8ZO3Zsmfynck1269aNBg0aMG/ePDZu3Fiiwli0XjTYT2VzCkv11LRpU66//npWr17N/fffX IHiiIffvghW7duPanjFXW9/vjjj09rOUWkrLDumXwZGA5MCh7LNO2YWRPguLsfMLO6wABgclxLKdWOmfGrX/2K2267DSjbKgmRLyaTJk1i7NixtG/fnkGDBtG2bVsOHTrERx99xLJly jTpw vvvoqAI8  iivvfYal1xyCe3atSMjI4ONGzeyYMECsrKyGDlyZFzPUaS422 /nQULFjBv3jwuvPBCBg0axOHDh5k9eza7d 9mzJgx9OnTJ xiSgLZfft4mj4yHoBCIgOT/WDMSN5s3YJ 1/yoTP7evXtz6623MnXqVM4//3yuvvrq6DyTWVlZZe6PPJVrMikpib59 0Z7ehSvTJ5zzjlkZ2fz4YcfkpycXO5AQFL9PfHEE2zevJl7772XmTNn0qdPH8466yx27NjBpk2bWLVqFS 88AJt27b92mPl5OSQlJTE2LFj2bBhA1lZkUkDSo QLSKnQaznHilvARoBucDm4LFhkN4ceCVYvwBYC6wHNgD3nuzxNc9k7Zafn 9JSUmelpbme/furTBfXl6eDx061Js1a YpKSneuHFjv/DCC3306NG atWqaL6FCxf6iBEj/Nxzz/XMzEyvV6 ed jQwW 99Vbftm1bPE5JapHy5l8rrmg veKOHDniDz74oHfu3NnT09M9IyPDe/fu7c8//3wcSizV0XsdO/nS71580vkLCwt96tSp3qlTJ09NTfVmzZr5zTff7AcOHCh3rr9TuSanTJnigGdmZvqJEydKbBs5cqQD3qNHj3L3Le/vokjR/MEzZsw46fOV2Kro/9zRo0d96tSp3qtXL8/MzPTU1FRv1aqV9 vXzx9//PEyn kVzTPp7j5z5ky/8MILPT09PTp3r0htQxzmmbTI69Qs3bt3d80nVXstXbqUnJwcbrjhhujIqyIi8pUDe/Kpd0Z9UtPTwi6KiIjEiJm97e5l50w6jcLq5ioSMw8//DAAt9xyS8glERFJTA2aNAy7CCIiUgOoMik1wrvvvsv8 fN5  23WbBgAZdffjkXXXRR2MUSEREREamxVJmUGuHtt9/md7/7HZmZmQwdOpRp06aFXSQRERERkRpN90yKiIiIiIjUMPG4ZzJm80yKiIiIiIhIzaXKpIiIiIiIiFSZKpMiIiIiIiJSZapMioiIiIiISJWpMikiIiIiIiJVpsqkiIiIiIiIVJkqkyIiIiIiIlJlNXKeSTPbA3xUKrkxsDeE4sjJUXwSm KT2BSfxKb4JDbFJ7EpPolN8UlsHd39jFi QJ1YHjws7t6kdJqZrY71pJ1y6hSfxKb4JDbFJ7EpPolN8Ulsik9iU3wSm5mtjvVrqJuriIiIiIiIVJkqkyIiIiIiIlJltakyOT3sAkilFJ/EpvgkNsUnsSk iU3xSWyKT2JTfBJbzONTIwfgERERERERkdiqTS2TIiIiIiIicprU Mqkmf3QzP5lZlvM7K6wy1OTmVkrM1tiZpvMbKOZ/b8gvaGZvW5mm4PHrGL7jA1i8y8zG1gsvZuZvRtsm2JmFqSnmdmLQfpKM2sT7/Oszsws2czWmtn84Llik0DMrIGZzTGz94O/o16KUWIws9HB/7UNZvaCmaUrNuEys6fNbLeZbSiWFpeYmNnw4DU2m9nw Jxx9VJBfP47 P 23sz z8waFNum MRRefEptu12M3Mza1wsTfGJo4riY2a3BjHYaGYPF0sPLz7uXmMXIBn4EGgHpALvAOeFXa6augDNgK7B hnAB8B5wMPAXUH6XcDkYP28ICZpQNsgVsnBtreAXoABC4DLgvSbgT8F69cCL4Z93tVpAW4DngfmB88VmwRagGeAnwfrqUADxSj8BWgBbAXqBs9nASMUm9Dj8n2gK7ChWFrMYwI0BP4dPGYF61lhvx JtlQQn0uBOsH6ZMUnseITpLcCFhKZr72x4pM48QFygEVAWvC8aSLEp6a3TPYAtrj7v939GPC/wOCQy1RjuftOd18TrH8ObCLyJWwwkS/JBI9DgvXBwP 6 1F33wpsAXqYWTMg091XeOTK/p9S xQdaw7Qv hXFqmcmbUEfgQ8VSxZsUkQZpZJ5MPjrwDufszdD6AYJYo6QF0zqwPUA3ag2ITK3d8A8kslxyMmA4HX3T3f3fcDrwM/PP1nWL2VFx93f83dTwRP3wRaBuuKT5xV8PcD8DgwBig qIriE2cVxOfXwCR3Pxrk2R2khxqfml6ZbAH8p9jz7UGaxFjQXP4dYCVwlrvvhEiFE2gaZKsoPi2C9dLpJfYJPpAOAo1icQ410B IfEAUFktTbBJHO2APMMMiXZGfMrP6KEahc/dPgEeAj4GdwEF3fw3FJhHFIyb6bnF63ESkpQQUn4RgZlcAn7j7O6U2KT6JoQNwcdAtdZmZfTdIDzU Nb0yWd6vuhq NsbMLAP4OzDK3T rLGs5aV5JemX7SCXM7HJgt7u/fbK7lJOm2MRWHSJdWp509 8AXxDpplcRxShOLHLf3WAi3YeaA/XN7IbKdiknTbEJ1 mMiWL1DZnZ3cAJ4LmipHKyKT5xZGb1gLuBe8vbXE6a4hN/dYh0Pe0J3AHMCloTQ41PTa9MbifS97tISyJdkyRGzCyFSEXyOXd/KUjeFTS1EzwWNctXFJ/tfNX1pXh6iX2C7mZnUn43DSmpN3CFmW0j0t27n5k9i2KTSLYD2919ZfB8DpHKpWIUvgHAVnff4 7HgZeA76HYJKJ4xETfLb6BYECPy4Hrg653oPgkgmwiP5i9E3xXaAmsMbOzUXwSxXbgJY94i0hPs8aEHJ aXplcBbQ3s7ZmlkrkBtOXQy5TjRX8OvJXYJO7P1Zs08tA0WhQw4F5xdKvDUaUagu0B94KuiZ9bmY9g2P tNQ Rce6Glhc7MNIKuDuY929pbu3IfJ3sNjdb0CxSRju/inwHzPrGCT1B95DMUoEHwM9zaxe8J72J3JPuGKTeOIRk4XApWaWFbRaXxqkydcwsx8CdwJXuPvhYpsUn5C5 7vu3tTd2wTfFbYTGVTxUxSfRDEX6AdgZh2IDNS3l7Dj4wkwYlEsF2AQkVFFPwTuDrs8NXkB hBpCl8PrAuWQUT6YOcCm4PHhsX2uTuIzb8IRpgK0rsDG4JtTwAWpKcDs4ncXPwW0C7s865uC3AJX43mqtgk0AJ0AVYHf0NziXRnUYwSYAHuA94P3teZREbNU2zCjckLRO5hPU7ki 9/xSsmRO732xIsPwv7vUjEpYL4bCFyP1bRd4Q/KT6JE59S27cRjOaq CRGfIhUHp8N3u81QL9EiE/RAUVEREREREROWk3v5ioiIiIiIiIxoMqkiIiIiIiIVJkqkyIiIiIiIlJlqkyKiIiIiIhIlakyKSIiIiIiIlWmyqSIiNR6Zna3mW00s/Vmts7MLorhay01s 6xOr6IiEi81Am7ACIiImEys17A5UQm6D5qZo2JzOclIiIilVDLpIiI1HbNgL3ufhTA3fe6 w4zu9fMVpnZBjObbmYG0ZbFx83sDTPbZGbfNbOXzGyzmT0Q5GljZu b2TNBa ccM6tX oXN7FIzW2Fma8xstpllBOmTzOy9YN9H4vheiIiInDRVJkVEpLZ7DWhlZh Y2TQz6xukP Hu33X384G6RFovixxz9 8DfwLmAb8BzgdGmFmjIE9HYLq7XwB8Btxc/EWDFtDfAwPcvSuwGrjNzBoCVwKdg30fiME5i4iIfGOqTIqISK3m7oeAbsBIYA/wopmNAHLMbKWZvQv0AzoX2 3l4PFdYKO77wxaNv8NtAq2/cfdlwfrzwJ9Sr10T A8YLmZrQOGA62JVDy/BJ4ys58Ah0/byYqIiJxGumdSRERqPXcvAJYCS4PK4y BC4Du7v4fMxsPpBfb5WjwWFhsveh50Werl36ZUs8NeN3drytdHjPrAfQHrgVuIVKZFRERSShqmRQRkVrNzDqaWftiSV2AfwXre4P7GK8 hUOfEwzuA3Ad8I9S298EepvZt4Jy1DOzDsHrnenurwCjgvKIiIgkHLVMiohIbZcBTDWzBsAJYAuRLq8HiHRj3QasOoXjbgKGm9mfgc3Ak8U3uvueoDvtC2aWFiT/HvgcmGdm6URaL0efwmuLiIjEnLmX7nUjIiIi34SZtQHmB4P3iIiI1Ejq5ioiIiIiIiJVppZJERERERERqTK1TIqIiIiIiEiVqTIpIiIiIiIiVabKpIiIiIiIiFSZKpMiIiIiIiJSZapMioiIiIiISJWpMikiIiIiIiJV9v8Bgk6P9/CgrY4AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 1080x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot\n",
    "plt.figure(figsize=(15, 5))\n",
    "plt.plot(myrec,label=\"Silence\")\n",
    "plt.xlabel(\"Samples\")\n",
    "plt.ylabel(\"Intensity\")\n",
    "plt.xlim(-1000, len(myrec))\n",
    "\n",
    "\n",
    "\n",
    "i=1\n",
    "splitted=[]\n",
    "chunk=[newPredictions[i]]\n",
    "while i <len(newPredictions):\n",
    "    if newPredictions[i]==newPredictions[i-1]:\n",
    "        chunk.append(newPredictions[i])\n",
    "    else:\n",
    "        splitted.append(chunk)\n",
    "        chunk=[newPredictions[i]]\n",
    "    i =1\n",
    "splitted.append(chunk)\n",
    "\n",
    "\n",
    "labels=[]\n",
    "yaxis=[]\n",
    "xaxis=[]\n",
    "start=step\n",
    "for j in range(len(splitted)):\n",
    "    k=len(splitted[j])\n",
    "    if splitted[j][0]!=-1:\n",
    "        labels.append(categories[splitted[j][0]])\n",
    "        yaxis.append(myrec[start:sr start step*(k-2)])\n",
    "        xaxis.append(list(range(start,sr start step*(k-2))))\n",
    "    start =step*k\n",
    "\n",
    "\n",
    "    \n",
    "for i in range(len(yaxis)):\n",
    "    plt.plot(xaxis[i],yaxis[i], label=labels[i])\n",
    "    #plt.vlines(xaxis[i][0], np.min(myrec), np.max(myrec), colors='k',linestyles='dashed')\n",
    "    #plt.vlines(xaxis[i][-1], np.min(myrec), np.max(myrec), colors='k',linestyles='dashed')\n",
    "    plt.text(xaxis[i][0]-step//2 (xaxis[i][-1]-xaxis[i][0])//2, np.min(myrec), labels[i], color='k',fontsize=20)\n",
    "\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "gputest",
   "language": "python",
   "name": "gputest"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}