ModelTuning_CNN_Model.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/Hani1-2/DeepLearningAssignmnt/blob/master/ModelTuning_CNN_Model.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "OamN8IBktMWV"
      },
      "source": [
        "## Load Data"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 2,
      "metadata": {
        "id": "1unMn2N0tQoC"
      },
      "outputs": [],
      "source": [
        "import json\n",
        "import numpy as np\n",
        "from sklearn.model_selection import train_test_split\n",
        "import matplotlib.pyplot as plt\n",
        "import tensorflow.keras as keras\n",
        "\n",
        "# path to json file that stores MFCCs and genre labels for each processed segment\n",
        "DATA_PATH = \"/content/drive/MyDrive/FYP/Modelling_folder/data_surah_1_spect_40.json\"\n",
        "\n",
        "def load_data(data_path):\n",
        "    \"\"\"Loads training dataset from json file.\n",
        "        :param data_path (str): Path to json file containing data\n",
        "        :return X (ndarray): Inputs\n",
        "        :return y (ndarray): Targets\n",
        "    \"\"\"\n",
        "\n",
        "    with open(data_path, \"r\") as fp:\n",
        "        data = json.load(fp)\n",
        "\n",
        "    # convert lists to numpy arrays\n",
        "    X = np.array(data[\"mfcc\"])\n",
        "    y = np.array(data[\"labels\"])\n",
        "\n",
        "    print(\"Data succesfully loaded!\")\n",
        "\n",
        "    return  X, y\n"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Z_W5crZ_DKQe",
        "outputId": "519b4d71-b04e-4f5a-c5a6-a13a54fa7e19"
      },
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Mounted at /content/drive\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "VwHjSamytRI5"
      },
      "source": [
        "# Plot History"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 3,
      "metadata": {
        "id": "oKL4Uj0gtU9I"
      },
      "outputs": [],
      "source": [
        "def plot_history(history):\n",
        "    \"\"\"Plots accuracy/loss for training/validation set as a function of the epochs\n",
        "        :param history: Training history of model\n",
        "        :return:\n",
        "    \"\"\"\n",
        "\n",
        "    fig, axs = plt.subplots(2)\n",
        "\n",
        "    # create accuracy sublpot\n",
        "    axs[0].plot(history.history[\"accuracy\"], label=\"train accuracy\")\n",
        "    axs[0].plot(history.history[\"val_accuracy\"], label=\"test accuracy\")\n",
        "    axs[0].set_ylabel(\"Accuracy\")\n",
        "    axs[0].legend(loc=\"lower right\")\n",
        "    axs[0].set_title(\"Accuracy eval\")\n",
        "\n",
        "    # create error sublpot\n",
        "    axs[1].plot(history.history[\"loss\"], label=\"train error\")\n",
        "    axs[1].plot(history.history[\"val_loss\"], label=\"test error\")\n",
        "    axs[1].set_ylabel(\"Error\")\n",
        "    axs[1].set_xlabel(\"Epoch\")\n",
        "    axs[1].legend(loc=\"upper right\")\n",
        "    axs[1].set_title(\"Error eval\")\n",
        "\n",
        "    plt.show()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "tOyNPDr89Etk"
      },
      "source": [
        "# Prepare Dataset"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 4,
      "metadata": {
        "id": "I98RRteJ3kCt"
      },
      "outputs": [],
      "source": [
        "def prepare_datasets(test_size, validation_size):\n",
        "    \"\"\"Loads data and splits it into train, validation and test sets.\n",
        "    :param test_size (float): Value in [0, 1] indicating percentage of data set to allocate to test split\n",
        "    :param validation_size (float): Value in [0, 1] indicating percentage of train set to allocate to validation split\n",
        "    :return X_train (ndarray): Input training set\n",
        "    :return X_validation (ndarray): Input validation set\n",
        "    :return X_test (ndarray): Input test set\n",
        "    :return y_train (ndarray): Target training set\n",
        "    :return y_validation (ndarray): Target validation set\n",
        "    :return y_test (ndarray): Target test set\n",
        "    \"\"\"\n",
        "\n",
        "    # load data\n",
        "    X, y = load_data(DATA_PATH)\n",
        "\n",
        "    # create train, validation and test split\n",
        "    X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=test_size)\n",
        "    X_train, X_validation, y_train, y_validation = train_test_split(X_train, y_train, test_size=validation_size)\n",
        "\n",
        "    # add an axis to input sets\n",
        "    X_train = X_train[..., np.newaxis]\n",
        "    X_validation = X_validation[..., np.newaxis]\n",
        "    X_test = X_test[..., np.newaxis]\n",
        "\n",
        "    return X_train, X_validation, X_test, y_train, y_validation, y_test\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "9T6b2ia29OGX"
      },
      "source": [
        "# Build Model"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 12,
      "metadata": {
        "id": "D4mdf_k36O_V"
      },
      "outputs": [],
      "source": [
        "def build_model(input_shape):\n",
        "    \"\"\"Generates CNN model\n",
        "    :param input_shape (tuple): Shape of input set\n",
        "    :return model: CNN model\n",
        "    \"\"\"\n",
        "\n",
        "    # build network topology\n",
        "    model = keras.Sequential()\n",
        "\n",
        "    # 1st conv layer\n",
        "    model.add(keras.layers.Conv2D(32, (2, 2), activation='relu', input_shape=input_shape))\n",
        "    model.add(keras.layers.MaxPooling2D((2, 2), strides=(2, 2), padding='same'))\n",
        "    model.add(keras.layers.Dropout(0.2))\n",
        "    # model.add(keras.layers.BatchNormalization())\n",
        "\n",
        "    # 2nd conv layer\n",
        "    model.add(keras.layers.Conv2D(64, (2, 2), activation='relu'))\n",
        "    model.add(keras.layers.MaxPooling2D((2, 2), strides=(2, 2), padding='same'))\n",
        "    model.add(keras.layers.Dropout(0.2))\n",
        "    # model.add(keras.layers.BatchNormalization())\n",
        "\n",
        "    # # 3rd conv layer\n",
        "    model.add(keras.layers.Conv2D(128, (2, 2), activation='relu'))\n",
        "    model.add(keras.layers.MaxPooling2D((2, 2), strides=(2, 2), padding='same'))\n",
        "    model.add(keras.layers.BatchNormalization())\n",
        "\n",
        "    # flatten output and feed it into dense layer\n",
        "    model.add(keras.layers.Flatten())\n",
        "    model.add(keras.layers.Dense(256, activation='relu'))\n",
        "    model.add(keras.layers.Dropout(0.2))\n",
        "\n",
        "    # output layer\n",
        "    model.add(keras.layers.Dense(28, activation='softmax'))\n",
        "\n",
        "    return model"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 13,
      "metadata": {
        "id": "XvcSJBv59S7z"
      },
      "outputs": [],
      "source": [
        "def predict(model, X, y):\n",
        "    \"\"\"Predict a single sample using the trained model\n",
        "    :param model: Trained classifier\n",
        "    :param X: Input data\n",
        "    :param y (int): Target\n",
        "    \"\"\"\n",
        "\n",
        "    # add a dimension to input data for sample - model.predict() expects a 4d array in this case\n",
        "    X = X[np.newaxis, ...] # array shape (1, 130, 13, 1)\n",
        "\n",
        "    # perform prediction\n",
        "    prediction = model.predict(X)\n",
        "\n",
        "    # get index with max value\n",
        "    predicted_index = np.argmax(prediction, axis=1)\n",
        "\n",
        "    print(\"Target: {}, Predicted label: {}\".format(y, predicted_index))\n"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# get train, validation, test splits\n",
        "X_train, X_validation, X_test, y_train, y_validation, y_test = prepare_datasets(0.25, 0.2)\n",
        "# create network\n",
        "input_shape = (X_train.shape[1], X_train.shape[2], 1)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "E-09yC_Uc0VO",
        "outputId": "a39fa07a-bf52-4c09-96f2-90cd61562304"
      },
      "execution_count": 14,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Data succesfully loaded!\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "X_train.shape"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "plsCBJJUWrUh",
        "outputId": "b2ef573d-970d-455d-bd83-fd8ab51bd66e"
      },
      "execution_count": 15,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(67248, 9, 40, 1)"
            ]
          },
          "metadata": {},
          "execution_count": 15
        }
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 16,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "SFPlVTH-9Vce",
        "outputId": "efeb7878-ba70-4712-ed92-798ae2bd9239"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(9, 40, 1)\n",
            "Model: \"sequential_1\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " conv2d_3 (Conv2D)           (None, 8, 39, 32)         160       \n",
            "                                                                 \n",
            " max_pooling2d_2 (MaxPooling  (None, 4, 20, 32)        0         \n",
            " 2D)                                                             \n",
            "                                                                 \n",
            " dropout_2 (Dropout)         (None, 4, 20, 32)         0         \n",
            "                                                                 \n",
            " conv2d_4 (Conv2D)           (None, 3, 19, 64)         8256      \n",
            "                                                                 \n",
            " max_pooling2d_3 (MaxPooling  (None, 2, 10, 64)        0         \n",
            " 2D)                                                             \n",
            "                                                                 \n",
            " dropout_3 (Dropout)         (None, 2, 10, 64)         0         \n",
            "                                                                 \n",
            " conv2d_5 (Conv2D)           (None, 1, 9, 128)         32896     \n",
            "                                                                 \n",
            " max_pooling2d_4 (MaxPooling  (None, 1, 5, 128)        0         \n",
            " 2D)                                                             \n",
            "                                                                 \n",
            " batch_normalization (BatchN  (None, 1, 5, 128)        512       \n",
            " ormalization)                                                   \n",
            "                                                                 \n",
            " flatten (Flatten)           (None, 640)               0         \n",
            "                                                                 \n",
            " dense (Dense)               (None, 256)               164096    \n",
            "                                                                 \n",
            " dropout_4 (Dropout)         (None, 256)               0         \n",
            "                                                                 \n",
            " dense_1 (Dense)             (None, 28)                7196      \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 213,116\n",
            "Trainable params: 212,860\n",
            "Non-trainable params: 256\n",
            "_________________________________________________________________\n",
            "Epoch 1/40\n",
            "2102/2102 [==============================] - 25s 6ms/step - loss: 2.4735 - accuracy: 0.2434 - val_loss: 1.5574 - val_accuracy: 0.4990\n",
            "Epoch 2/40\n",
            "2102/2102 [==============================] - 13s 6ms/step - loss: 1.4087 - accuracy: 0.5460 - val_loss: 0.6969 - val_accuracy: 0.7999\n",
            "Epoch 3/40\n",
            "2102/2102 [==============================] - 13s 6ms/step - loss: 0.8988 - accuracy: 0.7072 - val_loss: 0.4093 - val_accuracy: 0.8825\n",
            "Epoch 4/40\n",
            "2102/2102 [==============================] - 12s 6ms/step - loss: 0.6619 - accuracy: 0.7825 - val_loss: 0.2758 - val_accuracy: 0.9225\n",
            "Epoch 5/40\n",
            "2102/2102 [==============================] - 12s 6ms/step - loss: 0.5500 - accuracy: 0.8205 - val_loss: 0.2217 - val_accuracy: 0.9378\n",
            "Epoch 6/40\n",
            "2102/2102 [==============================] - 13s 6ms/step - loss: 0.4765 - accuracy: 0.8458 - val_loss: 0.1845 - val_accuracy: 0.9483\n",
            "Epoch 7/40\n",
            "2102/2102 [==============================] - 13s 6ms/step - loss: 0.4275 - accuracy: 0.8602 - val_loss: 0.1541 - val_accuracy: 0.9544\n",
            "Epoch 8/40\n",
            "2102/2102 [==============================] - 13s 6ms/step - loss: 0.3838 - accuracy: 0.8747 - val_loss: 0.1622 - val_accuracy: 0.9525\n",
            "Epoch 9/40\n",
            "2102/2102 [==============================] - 13s 6ms/step - loss: 0.3595 - accuracy: 0.8825 - val_loss: 0.1345 - val_accuracy: 0.9589\n",
            "Epoch 10/40\n",
            "2102/2102 [==============================] - 13s 6ms/step - loss: 0.3400 - accuracy: 0.8891 - val_loss: 0.1232 - val_accuracy: 0.9642\n",
            "Epoch 11/40\n",
            "2102/2102 [==============================] - 13s 6ms/step - loss: 0.3198 - accuracy: 0.8962 - val_loss: 0.1087 - val_accuracy: 0.9691\n",
            "Epoch 12/40\n",
            "2102/2102 [==============================] - 13s 6ms/step - loss: 0.3039 - accuracy: 0.9000 - val_loss: 0.1015 - val_accuracy: 0.9710\n",
            "Epoch 13/40\n",
            "2102/2102 [==============================] - 13s 6ms/step - loss: 0.2880 - accuracy: 0.9061 - val_loss: 0.0988 - val_accuracy: 0.9702\n",
            "Epoch 14/40\n",
            "2102/2102 [==============================] - 12s 6ms/step - loss: 0.2740 - accuracy: 0.9109 - val_loss: 0.0905 - val_accuracy: 0.9724\n",
            "Epoch 15/40\n",
            "2102/2102 [==============================] - 13s 6ms/step - loss: 0.2657 - accuracy: 0.9133 - val_loss: 0.0928 - val_accuracy: 0.9723\n",
            "Epoch 16/40\n",
            "2102/2102 [==============================] - 12s 6ms/step - loss: 0.2561 - accuracy: 0.9160 - val_loss: 0.0868 - val_accuracy: 0.9726\n",
            "Epoch 17/40\n",
            "2102/2102 [==============================] - 12s 6ms/step - loss: 0.2466 - accuracy: 0.9198 - val_loss: 0.0815 - val_accuracy: 0.9747\n",
            "Epoch 18/40\n",
            "2102/2102 [==============================] - 13s 6ms/step - loss: 0.2372 - accuracy: 0.9228 - val_loss: 0.0768 - val_accuracy: 0.9766\n",
            "Epoch 19/40\n",
            "2102/2102 [==============================] - 13s 6ms/step - loss: 0.2321 - accuracy: 0.9243 - val_loss: 0.0786 - val_accuracy: 0.9775\n",
            "Epoch 20/40\n",
            "2102/2102 [==============================] - 13s 6ms/step - loss: 0.2206 - accuracy: 0.9283 - val_loss: 0.0830 - val_accuracy: 0.9766\n",
            "Epoch 21/40\n",
            "2102/2102 [==============================] - 13s 6ms/step - loss: 0.2221 - accuracy: 0.9274 - val_loss: 0.0778 - val_accuracy: 0.9773\n",
            "Epoch 22/40\n",
            "2102/2102 [==============================] - 13s 6ms/step - loss: 0.2161 - accuracy: 0.9288 - val_loss: 0.0693 - val_accuracy: 0.9795\n",
            "Epoch 23/40\n",
            "2102/2102 [==============================] - 13s 6ms/step - loss: 0.2061 - accuracy: 0.9339 - val_loss: 0.0676 - val_accuracy: 0.9803\n",
            "Epoch 24/40\n",
            "2102/2102 [==============================] - 13s 6ms/step - loss: 0.2019 - accuracy: 0.9342 - val_loss: 0.0647 - val_accuracy: 0.9804\n",
            "Epoch 25/40\n",
            "2102/2102 [==============================] - 14s 6ms/step - loss: 0.1967 - accuracy: 0.9358 - val_loss: 0.0608 - val_accuracy: 0.9820\n",
            "Epoch 26/40\n",
            "2102/2102 [==============================] - 13s 6ms/step - loss: 0.1937 - accuracy: 0.9367 - val_loss: 0.0645 - val_accuracy: 0.9800\n",
            "Epoch 27/40\n",
            "2102/2102 [==============================] - 12s 6ms/step - loss: 0.1911 - accuracy: 0.9379 - val_loss: 0.0618 - val_accuracy: 0.9826\n",
            "Epoch 28/40\n",
            "2102/2102 [==============================] - 12s 6ms/step - loss: 0.1875 - accuracy: 0.9398 - val_loss: 0.0569 - val_accuracy: 0.9838\n",
            "Epoch 29/40\n",
            "2102/2102 [==============================] - 13s 6ms/step - loss: 0.1772 - accuracy: 0.9408 - val_loss: 0.0633 - val_accuracy: 0.9804\n",
            "Epoch 30/40\n",
            "2102/2102 [==============================] - 13s 6ms/step - loss: 0.1791 - accuracy: 0.9418 - val_loss: 0.0613 - val_accuracy: 0.9818\n",
            "Epoch 31/40\n",
            "2102/2102 [==============================] - 13s 6ms/step - loss: 0.1795 - accuracy: 0.9413 - val_loss: 0.0591 - val_accuracy: 0.9825\n",
            "Epoch 32/40\n",
            "2102/2102 [==============================] - 13s 6ms/step - loss: 0.1685 - accuracy: 0.9452 - val_loss: 0.0599 - val_accuracy: 0.9823\n",
            "Epoch 33/40\n",
            "2102/2102 [==============================] - 13s 6ms/step - loss: 0.1710 - accuracy: 0.9445 - val_loss: 0.0536 - val_accuracy: 0.9842\n",
            "Epoch 34/40\n",
            "2102/2102 [==============================] - 13s 6ms/step - loss: 0.1678 - accuracy: 0.9463 - val_loss: 0.0538 - val_accuracy: 0.9849\n",
            "Epoch 35/40\n",
            "2102/2102 [==============================] - 13s 6ms/step - loss: 0.1606 - accuracy: 0.9479 - val_loss: 0.0540 - val_accuracy: 0.9831\n",
            "Epoch 36/40\n",
            "2102/2102 [==============================] - 14s 6ms/step - loss: 0.1551 - accuracy: 0.9488 - val_loss: 0.0548 - val_accuracy: 0.9833\n",
            "Epoch 37/40\n",
            "2102/2102 [==============================] - 13s 6ms/step - loss: 0.1622 - accuracy: 0.9464 - val_loss: 0.0542 - val_accuracy: 0.9850\n",
            "Epoch 38/40\n",
            "2102/2102 [==============================] - 12s 6ms/step - loss: 0.1525 - accuracy: 0.9501 - val_loss: 0.0543 - val_accuracy: 0.9828\n",
            "Epoch 39/40\n",
            "2102/2102 [==============================] - 14s 6ms/step - loss: 0.1519 - accuracy: 0.9498 - val_loss: 0.0557 - val_accuracy: 0.9838\n",
            "Epoch 40/40\n",
            "2102/2102 [==============================] - 14s 6ms/step - loss: 0.1539 - accuracy: 0.9497 - val_loss: 0.0537 - val_accuracy: 0.9845\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 2 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "876/876 - 2s - loss: 0.0572 - accuracy: 0.9834 - 2s/epoch - 2ms/step\n",
            "\n",
            "Test accuracy: 0.9834052920341492\n",
            "1/1 [==============================] - 0s 181ms/step\n",
            "Target: 19, Predicted label: [19]\n"
          ]
        }
      ],
      "source": [
        "if __name__ == \"__main__\":\n",
        "\n",
        "    \n",
        "    print(input_shape)\n",
        "    model = build_model(input_shape)\n",
        "\n",
        "    # compile model\n",
        "    optimiser = keras.optimizers.Adam(learning_rate=0.001)\n",
        "    model.compile(optimizer=optimiser,\n",
        "                  loss='sparse_categorical_crossentropy',\n",
        "                  metrics=['accuracy'])\n",
        "\n",
        "    model.summary()\n",
        "\n",
        "    # train model\n",
        "    history = model.fit(X_train, y_train, validation_data=(X_validation, y_validation), batch_size=32, epochs=40)\n",
        "\n",
        "    # plot accuracy/error for training and validation\n",
        "    plot_history(history)\n",
        "\n",
        "    # evaluate model on test set\n",
        "    test_loss, test_acc = model.evaluate(X_test, y_test, verbose=2)\n",
        "    print('\\nTest accuracy:', test_acc)\n",
        "\n",
        "    # pick a sample to predict from the test set\n",
        "    X_to_predict = X_test[100]\n",
        "    y_to_predict = y_test[100]\n",
        "\n",
        "    # predict sample\n",
        "    predict(model, X_to_predict, y_to_predict)\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "BTD9uMSvNR9a"
      },
      "outputs": [],
      "source": [
        "SAVED_MODEL_PATH = \"model_spect_cnn.h5\"\n",
        "model.save(SAVED_MODEL_PATH)\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "OV5zC4cn9apU"
      },
      "outputs": [],
      "source": [
        "import pandas as pd"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "FfMiG8AIm7Zp"
      },
      "outputs": [],
      "source": [
        "_mapping = {\n",
        "    1:\"Bis'mi\",\n",
        "    2:\"Al-lahi\",\n",
        "    3:\"Al-rahmaani\",\n",
        "    4:\"Al-raheemi\",\n",
        "    5:\"Alhamdu\",\n",
        "    6:\"lillaahi\",\n",
        "    7:\"Rabbil\",\n",
        "    8:\"aalameen\",\n",
        "    9:\"Ar-Rahmaan\",\n",
        "    10:\"Ar-Raheem\",\n",
        "    11:\"Maaliki\",\n",
        "    12:\"Yumid\",\n",
        "    13:\"Diin\",\n",
        "    14:\"Iyyaka\",\n",
        "    15:\"Na'abudu\",\n",
        "    16:\"Iyyaka\",\n",
        "    17:\"Nasta'een\",\n",
        "    18:\"Ihdinas\",\n",
        "    19:\"Siraatal\",\n",
        "    20:\"Mustaqeem\",\n",
        "    21:\"Siraatal\",\n",
        "    22:\"Ladheena\",\n",
        "    23:\"An'amta\",\n",
        "    24:\"Alaihim\",\n",
        "    25:\"Ghayril\",\n",
        "    26:\"Maghdubi\",\n",
        "    27:\"Alaihim\",\n",
        "    28:\"Wala al-dalina\"}"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "Be0D-fewPUFg"
      },
      "outputs": [],
      "source": [
        "SAMPLE_RATE = 22050\n",
        "TRACK_DURATION = 2 # measured in seconds\n",
        "SAMPLES_PER_TRACK = SAMPLE_RATE * TRACK_DURATION\n",
        "file = '/content/003.wav'\n",
        "\n",
        "def save_mfcc(file_path,num_mfcc=13, n_fft=2048, hop_length=512, num_segments=10):\n",
        "    mfcc_lst = []\n",
        "    # we divide the track into 5 segments\n",
        "    # to calculate sample per segment - we need to know the number of samples per track (which is the sample rate * the track duration )\n",
        "    samples_per_segment = int(SAMPLES_PER_TRACK / num_segments)\n",
        "    num_mfcc_vectors_per_segment = math.ceil(samples_per_segment / hop_length)\n",
        "\n",
        "    signal, sample_rate = librosa.load(file_path, sr=SAMPLE_RATE)\n",
        "\n",
        "    # process all segments of audio file\n",
        "    for d in range(num_segments):\n",
        "\n",
        "        # calculate start and finish sample for current segment\n",
        "        start = samples_per_segment * d\n",
        "        finish = start + samples_per_segment\n",
        "\n",
        "        spec = librosa.stft(signal[start:finish], n_fft=2048, hop_length=512)\n",
        "        spec_power = librosa.power_to_db(np.abs(spec)**2, ref=np.max)\n",
        "\n",
        "        # extract mfcc - for each segment of signal\n",
        "        mfcc = librosa.feature.mfcc(S=spec_power, n_mfcc=13)\n",
        "        mfcc = mfcc.T\n",
        "        # store only mfcc feature with expected number of vectors\n",
        "        if len(mfcc) == num_mfcc_vectors_per_segment:\n",
        "            mfcc_lst.append(mfcc.tolist())\n",
        "    return mfcc_lst"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "mfc = save_mfcc('006.wav')\n",
        "x= mfc[0]\n",
        "y= mfc[1]\n",
        "print(len(mfc))\n",
        "# for i in mfc:\n",
        "  # print(f'Length of one segment extraction {len(i)}')\n",
        "\n",
        "print(len(mfc[0]))\n",
        "print(len(mfc[1][0]))\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "72K2E1kLec7W",
        "outputId": "91e7c98a-5061-4445-d4d0-a4a6e6195609"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "4\n",
            "9\n",
            "13\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.9/dist-packages/librosa/util/decorators.py:88: UserWarning: n_fft=2048 is too small for input signal of length=1472\n",
            "  return f(*args, **kwargs)\n",
            "/usr/local/lib/python3.9/dist-packages/librosa/util/decorators.py:88: UserWarning: n_fft=2048 is too small for input signal of length=0\n",
            "  return f(*args, **kwargs)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "x"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ab5jAriwiW1S",
        "outputId": "cac416de-1951-4afb-bf81-2be91357fc0e"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[[-2507.055908203125,\n",
              "  58.8436279296875,\n",
              "  42.574241638183594,\n",
              "  48.666358947753906,\n",
              "  42.18589782714844,\n",
              "  29.383399963378906,\n",
              "  31.5981388092041,\n",
              "  27.510051727294922,\n",
              "  17.401960372924805,\n",
              "  19.589622497558594,\n",
              "  16.579408645629883,\n",
              "  6.356538772583008,\n",
              "  9.928887367248535],\n",
              " [-2079.439453125,\n",
              "  384.4190979003906,\n",
              "  121.68710327148438,\n",
              "  199.00088500976562,\n",
              "  60.03031539916992,\n",
              "  -73.20272064208984,\n",
              "  67.14350891113281,\n",
              "  21.89228057861328,\n",
              "  -98.42532348632812,\n",
              "  31.946025848388672,\n",
              "  59.482242584228516,\n",
              "  -63.5679931640625,\n",
              "  -20.09283447265625],\n",
              " [-1867.736328125,\n",
              "  499.57501220703125,\n",
              "  53.550933837890625,\n",
              "  190.71197509765625,\n",
              "  74.42894744873047,\n",
              "  -81.75123596191406,\n",
              "  69.40164184570312,\n",
              "  -17.26337242126465,\n",
              "  -125.25604248046875,\n",
              "  75.84422302246094,\n",
              "  66.09994506835938,\n",
              "  -101.05113220214844,\n",
              "  -20.38248634338379],\n",
              " [-1811.3792724609375,\n",
              "  516.7086791992188,\n",
              "  5.573942184448242,\n",
              "  166.554931640625,\n",
              "  70.55339813232422,\n",
              "  -90.77044677734375,\n",
              "  66.41963195800781,\n",
              "  -21.082809448242188,\n",
              "  -127.74097442626953,\n",
              "  80.56095886230469,\n",
              "  55.699554443359375,\n",
              "  -122.05916595458984,\n",
              "  -25.589927673339844],\n",
              " [-1832.567626953125,\n",
              "  491.1020812988281,\n",
              "  -12.693096160888672,\n",
              "  161.07369995117188,\n",
              "  77.86074829101562,\n",
              "  -80.96417236328125,\n",
              "  90.51803588867188,\n",
              "  12.584928512573242,\n",
              "  -123.37271118164062,\n",
              "  69.72695922851562,\n",
              "  65.01394653320312,\n",
              "  -119.4219741821289,\n",
              "  -41.45520782470703],\n",
              " [-1950.853515625,\n",
              "  429.62078857421875,\n",
              "  24.125545501708984,\n",
              "  157.97296142578125,\n",
              "  67.57699584960938,\n",
              "  -72.69303894042969,\n",
              "  108.39774322509766,\n",
              "  75.68197631835938,\n",
              "  -79.8551254272461,\n",
              "  49.23120880126953,\n",
              "  70.5772705078125,\n",
              "  -89.50054931640625,\n",
              "  -45.585723876953125],\n",
              " [-2071.84814453125,\n",
              "  359.9066467285156,\n",
              "  43.39055252075195,\n",
              "  128.3402557373047,\n",
              "  41.87226867675781,\n",
              "  -64.4056396484375,\n",
              "  91.94569396972656,\n",
              "  84.06343078613281,\n",
              "  -37.826210021972656,\n",
              "  37.24862289428711,\n",
              "  49.74863815307617,\n",
              "  -40.42996597290039,\n",
              "  3.389554500579834],\n",
              " [-2068.259765625,\n",
              "  322.2603454589844,\n",
              "  43.51054382324219,\n",
              "  121.00279235839844,\n",
              "  25.85590362548828,\n",
              "  -52.94184112548828,\n",
              "  101.6606216430664,\n",
              "  78.39775085449219,\n",
              "  -43.26737976074219,\n",
              "  30.7987117767334,\n",
              "  53.21994400024414,\n",
              "  -22.455764770507812,\n",
              "  7.5745625495910645],\n",
              " [-1692.2034912109375,\n",
              "  215.72146606445312,\n",
              "  107.20035552978516,\n",
              "  83.54472351074219,\n",
              "  47.78977966308594,\n",
              "  33.323486328125,\n",
              "  49.141944885253906,\n",
              "  36.60014343261719,\n",
              "  22.721675872802734,\n",
              "  32.11030197143555,\n",
              "  18.50925064086914,\n",
              "  3.273447036743164,\n",
              "  15.62856674194336]]"
            ]
          },
          "metadata": {},
          "execution_count": 18
        }
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "q9U0v9aqn0HH",
        "outputId": "1446b145-93a1-42cf-9ace-54e6debdecbf"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.9/dist-packages/librosa/util/decorators.py:88: UserWarning: n_fft=2048 is too small for input signal of length=0\n",
            "  return f(*args, **kwargs)\n",
            "WARNING:tensorflow:6 out of the last 6 calls to <function Model.make_predict_function.<locals>.predict_function at 0x7f76d33623a0> triggered tf.function retracing. Tracing is expensive and the excessive number of tracings could be due to (1) creating @tf.function repeatedly in a loop, (2) passing tensors with different shapes, (3) passing Python objects instead of tensors. For (1), please define your @tf.function outside of the loop. For (2), @tf.function has reduce_retracing=True option that can avoid unnecessary retracing. For (3), please refer to https://www.tensorflow.org/guide/function#controlling_retracing and https://www.tensorflow.org/api_docs/python/tf/function for  more details.\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "1/1 [==============================] - 0s 101ms/step\n",
            "Alhamdu\n"
          ]
        }
      ],
      "source": [
        "\n",
        "import librosa\n",
        "import tensorflow as tf\n",
        "import numpy as np\n",
        "import math\n",
        "SAVED_MODEL_PATH = \"model_spect_cnn.h5\"\n",
        "SAMPLE_RATE = 22050\n",
        "TRACK_DURATION = 2 # measured in seconds\n",
        "SAMPLES_PER_TRACK = SAMPLE_RATE * TRACK_DURATION\n",
        "\n",
        "class _Keyword_Spotting_Service:\n",
        "    \"\"\"Singleton class for keyword spotting inference with trained models.\n",
        "\n",
        "    :param model: Trained model\n",
        "    \"\"\"\n",
        "    \n",
        "    model = None\n",
        "    _mapping = [\n",
        "    \"Bis'mi\",\n",
        "    \"Al-lahi\",\n",
        "    \"Al-rahmaani\",\n",
        "    \"Al-raheemi\",\n",
        "    \"Alhamdu\",\n",
        "    \"lillaahi\",\n",
        "    \"Rabbil\",\n",
        "    \"aalameen\",\n",
        "    \"Ar-Rahmaan\",\n",
        "    \"Ar-Raheem\",\n",
        "    \"Maaliki\",\n",
        "    \"Yumid\",\n",
        "    \"Diin\",\n",
        "    \"Iyyaka\",\n",
        "    \"Na'abudu\",\n",
        "    \"Iyyaka\",\n",
        "    \"Nasta'een\",\n",
        "    \"Ihdinas\",\n",
        "    \"Siraatal\",\n",
        "    \"Mustaqeem\",\n",
        "    \"Siraatal\",\n",
        "    \"Ladheena\",\n",
        "    \"An'amta\",\n",
        "    \"Alaihim\",\n",
        "    \"Ghayril\",\n",
        "    \"Maghdubi\",\n",
        "    \"Alaihim\",\n",
        "    \"Wala al-dalina\"\n",
        "    ]\n",
        "    _instance = None\n",
        "\n",
        "\n",
        "    def predict(self, file_path):\n",
        "        \"\"\"\n",
        "\n",
        "        :param file_path (str): Path to audio file to predict\n",
        "        :return predicted_keyword (str): Keyword predicted by the model\n",
        "        \"\"\"\n",
        "\n",
        "        # extract MFCC\n",
        "        # MFCCs = np.array([[-208.88369750976562, 173.29107666015625, -13.837435722351074, 14.191682815551758, -63.53321838378906, 9.908905029296875, -34.52302932739258, -7.0630083084106445, -11.072137832641602, 7.897786617279053, 12.635082244873047, -1.260940670967102, -14.440847396850586], [-175.98480224609375, 167.88204956054688, -30.289634704589844, 16.558849334716797, -79.59504699707031, 20.788333892822266, -40.25474548339844, -13.491707801818848, -13.442451477050781, 3.592095375061035, 10.45197868347168, -4.956630706787109, -13.133575439453125], [-177.2965545654297, 140.26446533203125, -42.05194091796875, 15.591264724731445, -79.16682434082031, 35.215572357177734, -48.15814971923828, -7.50286865234375, -7.240591049194336, 2.553539276123047, 9.138029098510742, -3.4423742294311523, -6.37964391708374], [-185.95509338378906, 124.90979766845703, -38.33936309814453, 22.440292358398438, -72.64540100097656, 47.9068717956543, -48.2619514465332, 1.8480926752090454, -3.999704360961914, -0.6501009464263916, 5.425509929656982, -6.084258079528809, -1.652677059173584], [-208.18551635742188, 112.87013244628906, -27.377513885498047, 31.67277717590332, -69.78677368164062, 45.79389953613281, -52.503211975097656, 0.9649603962898254, -7.219353199005127, -0.7659001350402832, 2.2723073959350586, -12.100591659545898, 0.217118501663208], [-190.88275146484375, 135.42083740234375, -31.51664924621582, 30.064231872558594, -69.34258270263672, 30.436262130737305, -56.083526611328125, -15.316060066223145, -16.722349166870117, 1.2809876203536987, -9.884056091308594, -20.502822875976562, -3.5870373249053955], [-207.81170654296875, 148.36880493164062, -37.037811279296875, 37.063602447509766, -56.117095947265625, 16.19550132751465, -35.421043395996094, -21.73065948486328, -18.389490127563477, 2.5057411193847656, -17.75062370300293, -18.73059844970703, -5.083065509796143], [-245.4512939453125, 137.54977416992188, -36.500579833984375, 54.88190841674805, -53.14923095703125, 13.605981826782227, -8.012560844421387, -28.28583526611328, -8.899959564208984, -3.0852625370025635, -13.78553581237793, -14.101425170898438, -10.775411605834961], [-236.1332244873047, 140.26885986328125, -4.87683629989624, 55.43954086303711, -48.025421142578125, 16.675363540649414, 1.0351996421813965, -34.397491455078125, -10.365550994873047, -0.5815303325653076, -10.888479232788086, -3.4156363010406494, -5.683342933654785]])\n",
        "        # MFCCs = np.array(save_mfcc(file_path))\n",
        "        mfc = save_mfcc(file_path)\n",
        "        x= mfc[0]\n",
        "        y= mfc[1]\n",
        "        MFCCs = np.array(y)\n",
        "        # we need a 4-dim array to feed to the model for prediction: (# samples, # time steps, # coefficients, 1)\n",
        "        MFCCs = MFCCs[np.newaxis, ..., np.newaxis]\n",
        "\n",
        "        # get the predicted label\n",
        "        predictions = self.model.predict(MFCCs) # a 2d array [[]]\n",
        "        predicted_index = np.argmax(predictions)\n",
        "        # index return the index which has highest score\n",
        "        predicted_keyword = self._mapping[predicted_index]\n",
        "        # print('prediction',predicted_index,predicted_keyword)\n",
        "        return predicted_keyword\n",
        "\n",
        "\n",
        "    def preprocess(self, file_path, num_mfcc=13, n_fft=2048, hop_length=512,num_segments=5):\n",
        "        mfcc_lst = []\n",
        "        # load audio file and slice it to ensure length consistency among different files\n",
        "        audio, sample_rate = librosa.load(file_path, res_type='kaiser_fast') \n",
        "\n",
        "        # Compute spectrogram\n",
        "        spec = librosa.stft(audio, n_fft=n_fft, hop_length=hop_length)\n",
        "\n",
        "        # Convert to power spectrogram\n",
        "        spec_power = librosa.power_to_db(np.abs(spec)**2, ref=np.max)\n",
        "\n",
        "        mfccs_features = librosa.feature.mfcc(S=spec_power, n_mfcc=13)\n",
        "        mfccs_scaled_features = np.mean(mfccs_features.T,axis=0)\n",
        "        mffccs = mfccs_scaled_features.tolist()\n",
        "        mfcc_lst.append(mffccs)\n",
        "        return mfcc_lst\n",
        "\n",
        "\n",
        "def Keyword_Spotting_Service():\n",
        "    \"\"\"Factory function for Keyword_Spotting_Service class.\n",
        "\n",
        "    :return _Keyword_Spotting_Service._instance (_Keyword_Spotting_Service):\n",
        "    \"\"\"\n",
        "\n",
        "    # ensure an instance is created only the first time the factory function is called\n",
        "    if _Keyword_Spotting_Service._instance is None:\n",
        "        _Keyword_Spotting_Service._instance = _Keyword_Spotting_Service()\n",
        "        _Keyword_Spotting_Service.model = tf.keras.models.load_model(SAVED_MODEL_PATH)\n",
        "    return _Keyword_Spotting_Service._instance\n",
        "\n",
        "\n",
        "if __name__ == \"__main__\":\n",
        "\n",
        "    # create 2 instances of the keyword spotting service\n",
        "    kss = Keyword_Spotting_Service()\n",
        "    kss1 = Keyword_Spotting_Service()\n",
        "\n",
        "    # check that different instances of the keyword spotting service point back to the same object (singleton)\n",
        "    assert kss is kss1\n",
        "\n",
        "    # make a prediction\n",
        "    keyword = kss.predict(\"380.wav\")\n",
        "    print(keyword)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "-B19MgJooI3F",
        "outputId": "6a8752d7-07bc-4c7a-f1ea-a9b2a06d19fd"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "13\n"
          ]
        }
      ],
      "source": [
        "import json\n",
        "with open(\"/content/drive/MyDrive/FYP-Code/data_surah_1_spect.json\") as fp:\n",
        "    data = json.load(fp)\n",
        "    print(len(data['mfcc'][0][0]))"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "WHjVoJ_k_27r",
        "outputId": "42c82ef3-8794-4731-aca4-0f8255760227"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "[[-1617.0831298828125, 356.81500244140625, 42.20360565185547, 127.14744567871094, 31.06903839111328, 37.66044616699219, 7.930259704589844, 69.84143829345703, 52.040443420410156, 8.817476272583008, 41.67535400390625, 38.18476104736328, 18.2706298828125], [-1344.205078125, 234.70086669921875, -72.42889404296875, 142.79278564453125, 34.36305236816406, 50.02061462402344, -21.355243682861328, 50.469810485839844, 31.13153076171875, 52.41215515136719, 43.89302062988281, 24.152965545654297, 1.8852910995483398], [-1229.3067626953125, 81.13493347167969, -148.42660522460938, 132.7469482421875, 19.93075942993164, 92.7179183959961, -24.369312286376953, 60.66669464111328, 15.434642791748047, 58.215538024902344, 39.374996185302734, 18.01659393310547, -3.4350860118865967], [-1198.5491943359375, 8.428567886352539, -183.18862915039062, 103.76251220703125, 6.750552177429199, 78.76371765136719, -31.91216468811035, 47.861270904541016, -1.6051545143127441, 42.336883544921875, 26.254680633544922, 23.33804702758789, -6.762016296386719], [-1190.636474609375, 22.389301300048828, -170.73422241210938, 106.75518798828125, 0.3240344524383545, 65.45597839355469, -34.216983795166016, 35.934051513671875, -22.98017120361328, 27.82012367248535, 14.872231483459473, 19.079790115356445, -20.077777862548828], [-1205.249267578125, 41.124053955078125, -176.50045776367188, 115.31864929199219, -11.080595016479492, 65.69047546386719, -25.39600372314453, 28.661319732666016, -15.979629516601562, 15.649227142333984, 11.941400527954102, 24.792476654052734, -15.729819297790527], [-1280.296875, 60.816158294677734, -152.48020935058594, 107.89934539794922, -2.1447253227233887, 80.95621490478516, -3.6147427558898926, 23.63401985168457, -17.21059226989746, -6.71693229675293, -2.0096733570098877, 28.13098907470703, 1.1635730266571045], [-1433.1529541015625, 146.5750732421875, -85.93475341796875, 133.07516479492188, 12.925132751464844, 102.44168090820312, 42.792396545410156, 56.703025817871094, -12.419278144836426, -3.6564841270446777, -11.404792785644531, 45.186134338378906, 36.469520568847656], [-1592.81787109375, 219.2447052001953, 80.10247802734375, 95.35132598876953, 60.85993957519531, 57.654205322265625, 67.17015838623047, 51.19758605957031, 24.42287826538086, 5.779135704040527, 6.419221878051758, 23.4683837890625, 16.560575485229492]]\n"
          ]
        }
      ],
      "source": [
        "print(data['mfcc'][0])"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "p3dC21aLSYAM",
        "outputId": "7b2bbabe-678b-44cb-c6f0-1ed39ce58519"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "13\n"
          ]
        }
      ],
      "source": [
        "import json\n",
        "with open(\"/content/drive/MyDrive/FYP-Code/data_spect_mfcc.json\") as fp:\n",
        "    data = json.load(fp)\n",
        "    print(len(data['MFCCs'][0]))"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "qWDNFfMYA49J",
        "outputId": "223bf5cc-1ddf-4790-dda8-c4213ea1fbbe"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Collecting librosa==0.9.1\n",
            "  Downloading librosa-0.9.1-py3-none-any.whl (213 kB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m213.1/213.1 KB\u001b[0m \u001b[31m8.1 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hRequirement already satisfied: packaging>=20.0 in /usr/local/lib/python3.9/dist-packages (from librosa==0.9.1) (23.0)\n",
            "Requirement already satisfied: scipy>=1.2.0 in /usr/local/lib/python3.9/dist-packages (from librosa==0.9.1) (1.10.1)\n",
            "Requirement already satisfied: numba>=0.45.1 in /usr/local/lib/python3.9/dist-packages (from librosa==0.9.1) (0.56.4)\n",
            "Requirement already satisfied: joblib>=0.14 in /usr/local/lib/python3.9/dist-packages (from librosa==0.9.1) (1.1.1)\n",
            "Requirement already satisfied: scikit-learn>=0.19.1 in /usr/local/lib/python3.9/dist-packages (from librosa==0.9.1) (1.2.2)\n",
            "Requirement already satisfied: soundfile>=0.10.2 in /usr/local/lib/python3.9/dist-packages (from librosa==0.9.1) (0.12.1)\n",
            "Requirement already satisfied: pooch>=1.0 in /usr/local/lib/python3.9/dist-packages (from librosa==0.9.1) (1.6.0)\n",
            "Requirement already satisfied: decorator>=4.0.10 in /usr/local/lib/python3.9/dist-packages (from librosa==0.9.1) (4.4.2)\n",
            "Requirement already satisfied: numpy>=1.17.0 in /usr/local/lib/python3.9/dist-packages (from librosa==0.9.1) (1.22.4)\n",
            "Requirement already satisfied: audioread>=2.1.5 in /usr/local/lib/python3.9/dist-packages (from librosa==0.9.1) (3.0.0)\n",
            "Collecting resampy>=0.2.2\n",
            "  Downloading resampy-0.4.2-py3-none-any.whl (3.1 MB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m3.1/3.1 MB\u001b[0m \u001b[31m73.5 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hRequirement already satisfied: llvmlite<0.40,>=0.39.0dev0 in /usr/local/lib/python3.9/dist-packages (from numba>=0.45.1->librosa==0.9.1) (0.39.1)\n",
            "Requirement already satisfied: setuptools in /usr/local/lib/python3.9/dist-packages (from numba>=0.45.1->librosa==0.9.1) (67.6.1)\n",
            "Requirement already satisfied: appdirs>=1.3.0 in /usr/local/lib/python3.9/dist-packages (from pooch>=1.0->librosa==0.9.1) (1.4.4)\n",
            "Requirement already satisfied: requests>=2.19.0 in /usr/local/lib/python3.9/dist-packages (from pooch>=1.0->librosa==0.9.1) (2.27.1)\n",
            "Requirement already satisfied: threadpoolctl>=2.0.0 in /usr/local/lib/python3.9/dist-packages (from scikit-learn>=0.19.1->librosa==0.9.1) (3.1.0)\n",
            "Requirement already satisfied: cffi>=1.0 in /usr/local/lib/python3.9/dist-packages (from soundfile>=0.10.2->librosa==0.9.1) (1.15.1)\n",
            "Requirement already satisfied: pycparser in /usr/local/lib/python3.9/dist-packages (from cffi>=1.0->soundfile>=0.10.2->librosa==0.9.1) (2.21)\n",
            "Requirement already satisfied: idna<4,>=2.5 in /usr/local/lib/python3.9/dist-packages (from requests>=2.19.0->pooch>=1.0->librosa==0.9.1) (3.4)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.9/dist-packages (from requests>=2.19.0->pooch>=1.0->librosa==0.9.1) (2022.12.7)\n",
            "Requirement already satisfied: urllib3<1.27,>=1.21.1 in /usr/local/lib/python3.9/dist-packages (from requests>=2.19.0->pooch>=1.0->librosa==0.9.1) (1.26.15)\n",
            "Requirement already satisfied: charset-normalizer~=2.0.0 in /usr/local/lib/python3.9/dist-packages (from requests>=2.19.0->pooch>=1.0->librosa==0.9.1) (2.0.12)\n",
            "Installing collected packages: resampy, librosa\n",
            "  Attempting uninstall: librosa\n",
            "    Found existing installation: librosa 0.10.0.post2\n",
            "    Uninstalling librosa-0.10.0.post2:\n",
            "      Successfully uninstalled librosa-0.10.0.post2\n",
            "Successfully installed librosa-0.9.1 resampy-0.4.2\n"
          ]
        }
      ],
      "source": [
        "!pip install librosa==0.9.1"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "aXkKUYUxA-BL"
      },
      "outputs": [],
      "source": [
        "import librosa\n",
        "import numpy as np"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "CM2o0HJWSwwO",
        "outputId": "aef9011e-1c38-463d-b174-3c7080e840a1"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "[[-2013.4505615234375, 387.5789794921875, -45.34129333496094, 28.32706642150879, 100.4990463256836, 14.555423736572266, 51.982200622558594, 56.059661865234375, -39.0953254699707, -9.492719650268555, 43.04741287231445, -12.159852027893066, -31.889345169067383]]\n"
          ]
        }
      ],
      "source": [
        "def preprocess(file_path, num_mfcc=13, n_fft=2048, hop_length=512,num_segments=5):\n",
        "        mfcc_lst = []\n",
        "        # load audio file and slice it to ensure length consistency among different files\n",
        "        audio, sample_rate = librosa.load(file_path, res_type='kaiser_fast') \n",
        "\n",
        "        # Compute spectrogram\n",
        "        spec = librosa.stft(audio, n_fft=n_fft, hop_length=hop_length)\n",
        "\n",
        "        # Convert to power spectrogram\n",
        "        spec_power = librosa.power_to_db(np.abs(spec)**2, ref=np.max)\n",
        "\n",
        "        mfccs_features = librosa.feature.mfcc(S=spec_power, n_mfcc=13)\n",
        "        mfccs_scaled_features = np.mean(mfccs_features.T,axis=0)\n",
        "        mffccs = mfccs_scaled_features.tolist()\n",
        "        mfcc_lst.append(mffccs)\n",
        "        return mfcc_lst\n",
        "print(preprocess('/content/002.wav'))"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "4uj-VsBXAqXy"
      },
      "outputs": [],
      "source": []
    }
  ],
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "provenance": [],
      "include_colab_link": true
    },
    "gpuClass": "standard",
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}