Speech Classification.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    },
    {
     "ename": "RuntimeError",
     "evalue": "module compiled against API version 0xc but this version of numpy is 0xa",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mRuntimeError\u001b[0m                              Traceback (most recent call last)",
      "\u001b[0;31mRuntimeError\u001b[0m: module compiled against API version 0xc but this version of numpy is 0xa"
     ]
    },
    {
     "ename": "RuntimeError",
     "evalue": "module compiled against API version 0xc but this version of numpy is 0xa",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mRuntimeError\u001b[0m                              Traceback (most recent call last)",
      "\u001b[0;31mRuntimeError\u001b[0m: module compiled against API version 0xc but this version of numpy is 0xa"
     ]
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "import matplotlib.pyplot as plt\n",
    "import os\n",
    "from scipy.io import wavfile\n",
    "from collections import defaultdict, Counter\n",
    "from scipy import signal\n",
    "import numpy as np\n",
    "import librosa\n",
    "import random as rn\n",
    "from keras.layers import Dense\n",
    "from keras import Input\n",
    "from keras.engine import Model\n",
    "from keras.utils import to_categorical\n",
    "from keras.layers import Dense, TimeDistributed, Dropout, Bidirectional, GRU, BatchNormalization, Activation, LeakyReLU, \\\n",
    "    LSTM, Flatten, RepeatVector, Permute, Multiply, Conv2D, MaxPooling2D"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "DATA_DIR = '/Users/dima/Projects/free-spoken-digit-dataset/recordings/'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "random_file = rn.choice(os.listdir(DATA_DIR))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sr: 22050\n",
      "wav shape: (10849,)\n",
      "length: 0.49201814059 secs\n"
     ]
    }
   ],
   "source": [
    "wav, sr = librosa.load(DATA_DIR + random_file)\n",
    "print 'sr:', sr\n",
    "print 'wav shape:', wav.shape\n",
    "print 'length:', wav.shape[0]/float(sr), 'secs'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sr: 8000\n",
      "wav shape: (1700,)\n",
      "length: 0.2125 secs\n"
     ]
    }
   ],
   "source": [
    "wav, sr = librosa.load(DATA_DIR + random_file, sr=None)\n",
    "print 'sr:', sr\n",
    "print 'wav shape:', wav.shape\n",
    "print 'length:', wav.shape[0]/float(sr), 'secs'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x12325de10>]"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYMAAAD8CAYAAACVZ8iyAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nhmEA/H8sbc+YGKFf0gAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x10d8ab250>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(wav)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x125aef7d0>"
      ]
     },
     "execution_count": 68,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZUAAADxCAYAAAADSSrDAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nDBkVIYQQNeP/Ac/qZHra6pGLAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x127858690>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "D = librosa.amplitude_to_db(np.abs(librosa.stft(wav)), ref=np.max)\n",
    "librosa.display.specshow(D, y_axis='linear')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train_X: (1000, 30000)\n",
      "train_spectrograms: (1000, 1025, 40)\n",
      "train_mel_spectrograms: (1000, 128, 40)\n",
      "train_mfccs: (1000, 20, 40)\n",
      "train_y: (1000, 10)\n",
      "\n",
      "test_X: (500, 30000)\n",
      "test_spectrograms: (500, 1025, 40)\n",
      "test_mel_spectrograms: (500, 128, 40)\n",
      "test_mfccs: (500, 20, 40)\n",
      "test_y: (500, 10)\n"
     ]
    }
   ],
   "source": [
    "test_speaker = 'theo'\n",
    "train_X = []\n",
    "train_spectrograms = []\n",
    "train_mel_spectrograms = []\n",
    "train_mfccs = []\n",
    "train_y = []\n",
    "\n",
    "test_X = []\n",
    "test_spectrograms = []\n",
    "test_mel_spectrograms = []\n",
    "test_mfccs = []\n",
    "test_y = []\n",
    "\n",
    "pad1d = lambda a, i: a[0: i] if a.shape[0] > i else np.hstack((a, np.zeros(i - a.shape[0])))\n",
    "pad2d = lambda a, i: a[:, 0: i] if a.shape[1] > i else np.hstack((a, np.zeros((a.shape[0],i - a.shape[1]))))\n",
    "\n",
    "for fname in os.listdir(DATA_DIR):\n",
    "    try:\n",
    "        if '.wav' not in fname or 'dima' in fname:\n",
    "            continue\n",
    "        struct = fname.split('_')\n",
    "        digit = struct[0]\n",
    "        speaker = struct[1]\n",
    "        wav, sr = librosa.load(DATA_DIR + fname)\n",
    "        padded_x = pad1d(wav, 30000)\n",
    "        spectrogram = np.abs(librosa.stft(wav))\n",
    "        padded_spectogram = pad2d(spectrogram,40)\n",
    "\n",
    "        mel_spectrogram = librosa.feature.melspectrogram(wav)\n",
    "        padded_mel_spectrogram = pad2d(mel_spectrogram,40)\n",
    "\n",
    "        mfcc = librosa.feature.mfcc(wav)\n",
    "        padded_mfcc = pad2d(mfcc,40)\n",
    "\n",
    "        if speaker == test_speaker:\n",
    "            test_X.append(padded_x)\n",
    "            test_spectrograms.append(padded_spectogram)\n",
    "            test_mel_spectrograms.append(padded_mel_spectrogram)\n",
    "            test_mfccs.append(padded_mfcc)\n",
    "            test_y.append(digit)\n",
    "        else:\n",
    "            train_X.append(padded_x)\n",
    "            train_spectrograms.append(padded_spectogram)\n",
    "            train_mel_spectrograms.append(padded_mel_spectrogram)\n",
    "            train_mfccs.append(padded_mfcc)\n",
    "            train_y.append(digit)\n",
    "    except Exception as e:\n",
    "        print fname, e\n",
    "        raise\n",
    "\n",
    "train_X = np.vstack(train_X)\n",
    "train_spectrograms = np.array(train_spectrograms)\n",
    "train_mel_spectrograms = np.array(train_mel_spectrograms)\n",
    "train_mfccs = np.array(train_mfccs)\n",
    "train_y = to_categorical(np.array(train_y))\n",
    "\n",
    "test_X = np.vstack(test_X)\n",
    "test_spectrograms = np.array(test_spectrograms)\n",
    "test_mel_spectrograms = np.array(test_mel_spectrograms)\n",
    "test_mfccs = np.array(test_mfccs)\n",
    "test_y = to_categorical(np.array(test_y))\n",
    "\n",
    "\n",
    "print 'train_X:', train_X.shape\n",
    "print 'train_spectrograms:', train_spectrograms.shape\n",
    "print 'train_mel_spectrograms:', train_mel_spectrograms.shape\n",
    "print 'train_mfccs:', train_mfccs.shape\n",
    "print 'train_y:', train_y.shape\n",
    "print\n",
    "print 'test_X:', test_X.shape\n",
    "print 'test_spectrograms:', test_spectrograms.shape\n",
    "print 'test_mel_spectrograms:', test_mel_spectrograms.shape\n",
    "print 'test_mfccs:', test_mfccs.shape\n",
    "print 'test_y:', test_y.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "ip = Input(shape=(train_X[0].shape))\n",
    "hidden = Dense(128, activation='relu')(ip)\n",
    "op = Dense(10, activation='softmax')(hidden)\n",
    "model = Model(input=ip, output=op)\n",
    "\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.compile(loss='categorical_crossentropy',\n",
    "              optimizer='adam',\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "history = model.fit(train_X,\n",
    "          train_y,\n",
    "          epochs=10,\n",
    "          batch_size=32,\n",
    "          validation_data=(test_X, test_y))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "plt.plot(history.history['acc'], label='Train Accuracy')\n",
    "plt.plot(history.history['val_acc'], label='Validation Accuracy')\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# SPECTROGRAMS CNN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train X shape: (1000, 1025, 40, 1)\n",
      "test X shape: (500, 1025, 40, 1)\n"
     ]
    }
   ],
   "source": [
    "train_X_ex = np.expand_dims(train_spectrograms, -1)\n",
    "test_X_ex = np.expand_dims(test_spectrograms, -1)\n",
    "print 'train X shape:', train_X_ex.shape\n",
    "print 'test X shape:', test_X_ex.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_2 (InputLayer)         (None, 1025, 40, 1)       0         \n",
      "_________________________________________________________________\n",
      "conv2d_4 (Conv2D)            (None, 1022, 37, 64)      1088      \n",
      "_________________________________________________________________\n",
      "max_pooling2d_4 (MaxPooling2 (None, 255, 9, 64)        0         \n",
      "_________________________________________________________________\n",
      "dropout_3 (Dropout)          (None, 255, 9, 64)        0         \n",
      "_________________________________________________________________\n",
      "flatten_2 (Flatten)          (None, 146880)            0         \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 32)                4700192   \n",
      "_________________________________________________________________\n",
      "dense_4 (Dense)              (None, 10)                330       \n",
      "=================================================================\n",
      "Total params: 4,701,610\n",
      "Trainable params: 4,701,610\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/dima/.virtualenvs/hs/lib/python2.7/site-packages/ipykernel/__main__.py:12: UserWarning: Update your `Model` call to the Keras 2 API: `Model(outputs=Tensor(\"de..., inputs=Tensor(\"in...)`\n"
     ]
    }
   ],
   "source": [
    "ip = Input(shape=train_X_ex[0].shape)\n",
    "m = Conv2D(32, kernel_size=(4, 4), activation='relu', padding='same')(ip)\n",
    "m = MaxPooling2D(pool_size=(4, 4))(m)\n",
    "m = Dropout(0.2)(m)\n",
    "m = Conv2D(64, kernel_size=(4, 4), activation='relu')(ip)\n",
    "m = MaxPooling2D(pool_size=(4, 4))(m)\n",
    "m = Dropout(0.2)(m)\n",
    "m = Flatten()(m)\n",
    "m = Dense(32, activation='relu')(m)\n",
    "op = Dense(10, activation='softmax')(m)\n",
    "\n",
    "model = Model(input=ip, output=op)\n",
    "\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 1000 samples, validate on 500 samples\n",
      "Epoch 1/10\n",
      "1000/1000 [==============================] - 37s 37ms/step - loss: 1.2106 - acc: 0.6540 - val_loss: 1.8451 - val_acc: 0.3500\n",
      "Epoch 2/10\n",
      "1000/1000 [==============================] - 36s 36ms/step - loss: 0.2179 - acc: 0.9360 - val_loss: 1.7112 - val_acc: 0.5440\n",
      "Epoch 3/10\n",
      "1000/1000 [==============================] - 35s 35ms/step - loss: 0.1119 - acc: 0.9670 - val_loss: 1.5084 - val_acc: 0.6500\n",
      "Epoch 4/10\n",
      "1000/1000 [==============================] - 36s 36ms/step - loss: 0.0486 - acc: 0.9860 - val_loss: 1.5877 - val_acc: 0.5100\n",
      "Epoch 5/10\n",
      "1000/1000 [==============================] - 36s 36ms/step - loss: 0.0523 - acc: 0.9810 - val_loss: 1.4572 - val_acc: 0.6680\n",
      "Epoch 6/10\n",
      "1000/1000 [==============================] - 36s 36ms/step - loss: 0.0223 - acc: 0.9960 - val_loss: 1.4729 - val_acc: 0.6420\n",
      "Epoch 7/10\n",
      "1000/1000 [==============================] - 36s 36ms/step - loss: 0.0269 - acc: 0.9920 - val_loss: 1.4190 - val_acc: 0.6420\n",
      "Epoch 8/10\n",
      "1000/1000 [==============================] - 36s 36ms/step - loss: 0.0787 - acc: 0.9830 - val_loss: 1.4275 - val_acc: 0.6480\n",
      "Epoch 9/10\n",
      "1000/1000 [==============================] - 36s 36ms/step - loss: 0.0556 - acc: 0.9890 - val_loss: 1.4976 - val_acc: 0.5940\n",
      "Epoch 10/10\n",
      "1000/1000 [==============================] - 36s 36ms/step - loss: 0.0133 - acc: 0.9980 - val_loss: 1.4833 - val_acc: 0.6160\n"
     ]
    }
   ],
   "source": [
    "model.compile(loss='categorical_crossentropy',\n",
    "              optimizer='adam',\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "history = model.fit(train_X_ex,\n",
    "          train_y,\n",
    "          epochs=10,\n",
    "          batch_size=32,\n",
    "          verbose=1,\n",
    "          validation_data=(test_X_ex, test_y))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x12359a910>"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEKCAYAAAD9xUlFAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nNSMjnQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x10e546550>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(history.history['acc'], label='Train Accuracy')\n",
    "plt.plot(history.history['val_acc'], label='Validation Accuracy')\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# MEL SPECTROGRAM CNN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train X shape: (1000, 128, 40, 1)\n",
      "test X shape: (500, 128, 40, 1)\n"
     ]
    }
   ],
   "source": [
    "train_X_ex = np.expand_dims(train_mel_spectrograms, -1)\n",
    "test_X_ex = np.expand_dims(test_mel_spectrograms, -1)\n",
    "print 'train X shape:', train_X_ex.shape\n",
    "print 'test X shape:', test_X_ex.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_3 (InputLayer)         (None, 128, 40, 1)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_6 (Conv2D)            (None, 125, 37, 64)       1088      \n",
      "_________________________________________________________________\n",
      "max_pooling2d_6 (MaxPooling2 (None, 31, 9, 64)         0         \n",
      "_________________________________________________________________\n",
      "dropout_5 (Dropout)          (None, 31, 9, 64)         0         \n",
      "_________________________________________________________________\n",
      "flatten_3 (Flatten)          (None, 17856)             0         \n",
      "_________________________________________________________________\n",
      "dense_5 (Dense)              (None, 32)                571424    \n",
      "_________________________________________________________________\n",
      "dense_6 (Dense)              (None, 10)                330       \n",
      "=================================================================\n",
      "Total params: 572,842\n",
      "Trainable params: 572,842\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/dima/.virtualenvs/hs/lib/python2.7/site-packages/ipykernel/__main__.py:12: UserWarning: Update your `Model` call to the Keras 2 API: `Model(outputs=Tensor(\"de..., inputs=Tensor(\"in...)`\n"
     ]
    }
   ],
   "source": [
    "ip = Input(shape=train_X_ex[0].shape)\n",
    "m = Conv2D(32, kernel_size=(4, 4), activation='relu', padding='same')(ip)\n",
    "m = MaxPooling2D(pool_size=(4, 4))(m)\n",
    "m = Dropout(0.2)(m)\n",
    "m = Conv2D(64, kernel_size=(4, 4), activation='relu')(ip)\n",
    "m = MaxPooling2D(pool_size=(4, 4))(m)\n",
    "m = Dropout(0.2)(m)\n",
    "m = Flatten()(m)\n",
    "m = Dense(32, activation='relu')(m)\n",
    "op = Dense(10, activation='softmax')(m)\n",
    "\n",
    "model = Model(input=ip, output=op)\n",
    "\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 1000 samples, validate on 500 samples\n",
      "Epoch 1/10\n",
      "1000/1000 [==============================] - 5s 5ms/step - loss: 2.1539 - acc: 0.4050 - val_loss: 2.2378 - val_acc: 0.2060\n",
      "Epoch 2/10\n",
      "1000/1000 [==============================] - 4s 4ms/step - loss: 1.0881 - acc: 0.6450 - val_loss: 2.2974 - val_acc: 0.1640\n",
      "Epoch 3/10\n",
      "1000/1000 [==============================] - 4s 4ms/step - loss: 0.7304 - acc: 0.8020 - val_loss: 2.4219 - val_acc: 0.1600\n",
      "Epoch 4/10\n",
      "1000/1000 [==============================] - 4s 4ms/step - loss: 0.5459 - acc: 0.8530 - val_loss: 2.4777 - val_acc: 0.1580\n",
      "Epoch 5/10\n",
      "1000/1000 [==============================] - 4s 4ms/step - loss: 0.3867 - acc: 0.9170 - val_loss: 2.6358 - val_acc: 0.1660\n",
      "Epoch 6/10\n",
      "1000/1000 [==============================] - 4s 4ms/step - loss: 0.3840 - acc: 0.9110 - val_loss: 2.5003 - val_acc: 0.1800\n",
      "Epoch 7/10\n",
      "1000/1000 [==============================] - 4s 4ms/step - loss: 0.3243 - acc: 0.9190 - val_loss: 2.5728 - val_acc: 0.2120\n",
      "Epoch 8/10\n",
      "1000/1000 [==============================] - 4s 4ms/step - loss: 0.3057 - acc: 0.9290 - val_loss: 2.5160 - val_acc: 0.2220\n",
      "Epoch 9/10\n",
      "1000/1000 [==============================] - 4s 4ms/step - loss: 0.3020 - acc: 0.9330 - val_loss: 2.7102 - val_acc: 0.2660\n",
      "Epoch 10/10\n",
      "1000/1000 [==============================] - 4s 4ms/step - loss: 0.2579 - acc: 0.9360 - val_loss: 2.6487 - val_acc: 0.1880\n"
     ]
    }
   ],
   "source": [
    "model.compile(loss='categorical_crossentropy',\n",
    "              optimizer='adam',\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "history = model.fit(train_X_ex,\n",
    "          train_y,\n",
    "          epochs=10,\n",
    "          batch_size=32,\n",
    "          verbose=1,\n",
    "          validation_data=(test_X_ex, test_y))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x1232d1510>"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEKCAYAAAD9xUlFAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x10e546ad0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(history.history['acc'], label='Train Accuracy')\n",
    "plt.plot(history.history['val_acc'], label='Validation Accuracy')\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# MFCC CNN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_X_ex = np.expand_dims(train_mfccs, -1)\n",
    "test_X_ex = np.expand_dims(test_mfccs, -1)\n",
    "print 'train X shape:', train_X_ex.shape\n",
    "print 'test X shape:', test_X_ex.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "ip = Input(shape=train_X_ex[0].shape)\n",
    "m = Conv2D(64, kernel_size=(4, 4), activation='relu')(ip)\n",
    "m = MaxPooling2D(pool_size=(4, 4))(m)\n",
    "# m = Conv2D(128, kernel_size=(2, 2), activation='relu')(ip)\n",
    "# m = MaxPooling2D(pool_size=(2, 2))(m)\n",
    "m = Flatten()(m)\n",
    "m = Dense(32, activation='relu')(m)\n",
    "op = Dense(10, activation='softmax')(m)\n",
    "\n",
    "model = Model(input=ip, output=op)\n",
    "\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.compile(loss='categorical_crossentropy',\n",
    "              optimizer='adam',\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "history = model.fit(train_X_ex,\n",
    "          train_y,\n",
    "          epochs=100,\n",
    "          batch_size=32,\n",
    "          verbose=0,\n",
    "          validation_data=(test_X_ex, test_y))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "plt.plot(history.history['acc'], label='Train Accuracy')\n",
    "plt.plot(history.history['val_acc'], label='Validation Accuracy')\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.legend()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python (hiredscore)",
   "language": "python",
   "name": "hiredscore"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}