intialCommit

.ipynb_checkpoints/music-genre-classification-checkpoint.ipynb

@@ -0,0 +1,279 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Music Genre Classification"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "metadata": {
+    "pycharm": {
+     "is_executing": true
+    }
+   },
+   "outputs": [],
+   "source": [
+    "from python_speech_features import mfcc\n",
+    "import scipy.io.wavfile as wav\n",
+    "import numpy as np\n",
+    "\n",
+    "from tempfile import TemporaryFile\n",
+    "\n",
+    "import os\n",
+    "import pickle\n",
+    "import random\n",
+    "import operator\n",
+    "\n",
+    "import math"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# function to get the distance between feature vecotrs and find neighbors\n",
+    "def getNeighbors(trainingSet, instance, k):\n",
+    "    distances = []\n",
+    "    for x in range(len(trainingSet)):\n",
+    "        dist = distance(trainingSet[x], instance, k) + distance(instance, trainingSet[x], k)\n",
+    "        distances.append((trainingSet[x][2], dist))\n",
+    "\n",
+    "    distances.sort(key=operator.itemgetter(1))\n",
+    "    neighbors = []\n",
+    "    for x in range(k):\n",
+    "        neighbors.append(distances[x][0])\n",
+    "    \n",
+    "    return neighbors"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# identify the class of the instance\n",
+    "def nearestClass(neighbors):\n",
+    "    classVote = {}\n",
+    "\n",
+    "    for x in range(len(neighbors)):\n",
+    "        response = neighbors[x]\n",
+    "        if response in classVote:\n",
+    "            classVote[response] += 1\n",
+    "        else:\n",
+    "            classVote[response] = 1\n",
+    "\n",
+    "    sorter = sorted(classVote.items(), key = operator.itemgetter(1), reverse=True)\n",
+    "\n",
+    "    return sorter[0][0]"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# function to evaluate the model\n",
+    "def getAccuracy(testSet, prediction):\n",
+    "    correct = 0\n",
+    "    for x in range(len(testSet)):\n",
+    "        if testSet[x][-1] == predictions[x]:\n",
+    "            correct += 1\n",
+    "    \n",
+    "    return (1.0 * correct) / len(testSet)\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 6,
+   "metadata": {
+    "tags": []
+   },
+   "outputs": [],
+   "source": [
+    "# directory that holds the wav files\n",
+    "directory = \"Data/genres_original/\"\n",
+    "\n",
+    "# binary file where we will collect all the features extracted using mfcc (Mel Frequency Cepstral Coefficients)\n",
+    "f = open(\"my.dat\", 'wb')\n",
+    "\n",
+    "i = 0\n",
+    "\n",
+    "for folder in os.listdir(directory):\n",
+    "    i += 1\n",
+    "    if i == 11:\n",
+    "        break\n",
+    "    for file in os.listdir(directory+folder):        \n",
+    "        try:\n",
+    "            (rate, sig) = wav.read(directory+folder+\"/\"+file)\n",
+    "            mfcc_feat = mfcc(sig, rate, winlen=0.020, appendEnergy=False)\n",
+    "            covariance = np.cov(np.matrix.transpose(mfcc_feat))\n",
+    "            mean_matrix = mfcc_feat.mean(0)\n",
+    "            feature = (mean_matrix, covariance, i)\n",
+    "            pickle.dump(feature, f)\n",
+    "        except Exception as e:\n",
+    "            print('Got an exception: ', e, ' in folder: ', folder, ' filename: ', file)        \n",
+    "\n",
+    "f.close()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 7,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Split the dataset into training and testing sets respectively\n",
+    "dataset = []\n",
+    "\n",
+    "def loadDataset(filename, split, trSet, teSet):\n",
+    "    with open('my.dat', 'rb') as f:\n",
+    "        while True:\n",
+    "            try:\n",
+    "                dataset.append(pickle.load(f))\n",
+    "            except EOFError:\n",
+    "                f.close()\n",
+    "                break\n",
+    "    for x in range(len(dataset)):\n",
+    "        if random.random() < split:\n",
+    "            trSet.append(dataset[x])\n",
+    "        else:\n",
+    "            teSet.append(dataset[x])\n",
+    "trainingSet = []\n",
+    "testSet = []\n",
+    "loadDataset('my.dat', 0.66, trainingSet, testSet)\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 8,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "def distance(instance1 , instance2 , k ):\n",
+    "    distance =0 \n",
+    "    mm1 = instance1[0] \n",
+    "    cm1 = instance1[1]\n",
+    "    mm2 = instance2[0]\n",
+    "    cm2 = instance2[1]\n",
+    "    distance = np.trace(np.dot(np.linalg.inv(cm2), cm1)) \n",
+    "    distance+=(np.dot(np.dot((mm2-mm1).transpose() , np.linalg.inv(cm2)) , mm2-mm1 )) \n",
+    "    distance+= np.log(np.linalg.det(cm2)) - np.log(np.linalg.det(cm1))\n",
+    "    distance-= k\n",
+    "    return distance"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "tags": []
+   },
+   "outputs": [],
+   "source": [
+    "# making predictions using KNN\n",
+    "leng = len(testSet)\n",
+    "predictions = []\n",
+    "for x in range(leng):\n",
+    "    predictions.append(nearestClass(getNeighbors(trainingSet, testSet[x], 5)))\n",
+    "\n",
+    "accuracy1 = getAccuracy(testSet, predictions)\n",
+    "print(accuracy1)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# testing the code with external samples\n",
+    "# URL: https://uweb.engr.arizona.edu/~429rns/audiofiles/audiofiles.html\n",
+    "\n",
+    "test_dir = \"Test/\"\n",
+    "# test_file = test_dir + \"test.wav\"\n",
+    "test_file = test_dir + \"3.wav\"\n",
+    "# test_file = test_dir + \"test4.wav\""
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "tags": []
+   },
+   "outputs": [],
+   "source": [
+    "(rate, sig) = wav.read(test_file)\n",
+    "mfcc_feat = mfcc(sig, rate, winlen=0.020, appendEnergy=False)\n",
+    "covariance = np.cov(np.matrix.transpose(mfcc_feat))\n",
+    "mean_matrix = mfcc_feat.mean(0)\n",
+    "feature = (mean_matrix, covariance, i)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from collections import defaultdict\n",
+    "results = defaultdict(int)\n",
+    "\n",
+    "directory = \"Data/genres_original/\"\n",
+    "\n",
+    "i = 1\n",
+    "for folder in os.listdir(directory):\n",
+    "    results[i] = folder\n",
+    "    i += 1\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "tags": []
+   },
+   "outputs": [],
+   "source": [
+    "pred = nearestClass(getNeighbors(dataset, feature, 5))\n",
+    "print(results[pred])"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3 (ipykernel)",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.9.7"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}