Test.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Test MFCC"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# of test:  11005\n",
      "# of val:  9981\n",
      "# of train:  84849\n",
      "# total:  105835\n",
      "(10000, 16000)\n",
      "Loading  10000  files (seconds): 4.3908851146698\n",
      "Size: 1280000000 bytes\n",
      "Transforming  10000  elements (seconds): 41.128499031066895\n",
      "Size: 362880000 bytes\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import librosa\n",
    "\n",
    "import pandas as pd # for reading txt files\n",
    "import time # for measuring time\n",
    "import os\n",
    "\n",
    "sr=16000\n",
    "\n",
    "\n",
    "basePath = 'sd_GSCmdV2'\n",
    "testWAVs = pd.read_csv(basePath + '/train/testing_list.txt',\n",
    "                           sep=\" \", header=None)[0].tolist()\n",
    "testWAVs = [os.path.join(basePath + '/train/', f + '.npy')\n",
    "                for f in testWAVs if f.endswith('.wav')]\n",
    "\n",
    "valWAVs = pd.read_csv(basePath + '/train/validation_list.txt',\n",
    "                          sep=\" \", header=None)[0].tolist()\n",
    "valWAVs = [os.path.join(basePath + '/train/', f + '.npy')\n",
    "               for f in valWAVs if f.endswith('.wav')]\n",
    "allWAVs = []\n",
    "for root, dirs, files in os.walk(basePath + '/train/'):\n",
    "    allWAVs += [root + '/' + f for f in files if f.endswith('.wav.npy')]\n",
    "trainWAVs = list(set(allWAVs) - set(valWAVs) - set(testWAVs))\n",
    "\n",
    "print(\"# of test: \",len(testWAVs))\n",
    "print(\"# of val: \",len(valWAVs))\n",
    "print(\"# of train: \",len(trainWAVs))\n",
    "print(\"# total: \",len(allWAVs))\n",
    "\n",
    "\n",
    "#LOAD A BATCH OF DATA\n",
    "batch_size=10000\n",
    "dim=sr\n",
    "start = time.time()\n",
    "X = np.empty((batch_size, dim))\n",
    "print(X.shape)\n",
    "# Generate data\n",
    "for i, ID in enumerate(allWAVs[0:batch_size]):\n",
    "    # load data from file, saved as numpy array on disk\n",
    "    curX = np.load(ID)\n",
    "\n",
    "    # curX could be bigger or smaller than self.dim\n",
    "    if curX.shape[0] == dim:\n",
    "        X[i] = curX\n",
    "    elif curX.shape[0] > dim:  # bigger\n",
    "        # we can choose any position in curX-self.dim\n",
    "        randPos = np.random.randint(curX.shape[0]-dim)\n",
    "        X[i] = curX[randPos:randPos+dim]\n",
    "    else:  # smaller\n",
    "        randPos = np.random.randint(dim-curX.shape[0])\n",
    "        X[i, randPos:randPos + curX.shape[0]] = curX\n",
    "        # print('File dim smaller')\n",
    "\n",
    "end = time.time()\n",
    "print(\"Loading \", batch_size, \" files (seconds):\", end - start)\n",
    "print(\"Size: %d bytes\" % (X.nbytes))\n",
    "\n",
    "\n",
    "# FEATURE EXTRACTION OF A BATCH\n",
    "start = time.time()\n",
    "features = np.empty((batch_size,36,126))\n",
    "for i,y in enumerate(X):\n",
    "    S = librosa.feature.melspectrogram(y, sr=sr, n_fft=1024,\n",
    "                                            hop_length=128, power=1.0, #window='hann',\n",
    "                                            n_mels=80, fmin=40.0, fmax=sr/2)\n",
    "\n",
    "    # Convert to log scale (dB). We'll use the peak power (max) as reference.\n",
    "    log_S = librosa.power_to_db(S, ref=np.max)\n",
    "\n",
    "    # Next, we'll extract the top 13 Mel-frequency cepstral coefficients (MFCCs)\n",
    "    mfcc        = librosa.feature.mfcc(S=log_S, n_mfcc=12)\n",
    "\n",
    "    # Let's pad on the first and second deltas while we're at it\n",
    "    delta_mfcc  = librosa.feature.delta(mfcc)\n",
    "    delta2_mfcc = librosa.feature.delta(mfcc, order=2)\n",
    "\n",
    "    features[i] = np.concatenate((mfcc, delta_mfcc, delta2_mfcc), axis=0)\n",
    "\n",
    "end = time.time()\n",
    "print(\"Transforming \", batch_size, \" elements (seconds):\", end - start)\n",
    "print(\"Size: %d bytes\" % (features.nbytes))\n",
    "    \n",
    "    \n",
    "'''plt.figure(figsize=(10, 4))\n",
    "plt.pcolormesh(mfcc)\n",
    "plt.show()\n",
    "\n",
    "plt.figure(figsize=(10, 4))\n",
    "plt.pcolormesh(delta_mfcc)\n",
    "plt.show()\n",
    "\n",
    "plt.figure(figsize=(10, 4))\n",
    "plt.pcolormesh(delta2_mfcc)\n",
    "plt.show()'''\n",
    "\n",
    "plt.figure(figsize=(10, 4))\n",
    "plt.pcolormesh(features[0])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "del(X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "%reset_selective -f X"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "%reset_selective -f features"
   ]
  },
  {
   "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
}