data_preprocess.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "id": "85czbYW8yqdu"
   },
   "outputs": [],
   "source": [
    "import librosa\n",
    "import soundfile as sf\n",
    "\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "\n",
    "import os\n",
    "from os import listdir\n",
    "from os.path import isfile, join\n",
    "\n",
    "from collections import OrderedDict\n",
    "\n",
    "import json"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "id": "aoO0u8wU0D9p"
   },
   "outputs": [],
   "source": [
    "def extract_features(filepath):\n",
    "    \n",
    "    '''\n",
    "    Source: https://github.com/danz1ka19/Music-Emotion-Recognition/blob/master/Feature-Extraction.py\n",
    "    Modified to process a single file\n",
    "\n",
    "        function: extract_features\n",
    "        input: path to mp3 files\n",
    "        output: csv file containing features extracted\n",
    "\n",
    "        This function reads the content in a directory and for each audio file detected\n",
    "        reads the file and extracts relevant features using librosa library for audio\n",
    "        signal processing\n",
    "    '''\n",
    "\n",
    "    feature_set = {}  # Features\n",
    "\n",
    "    # Reading audio file\n",
    "    y, sr = librosa.load(filepath)\n",
    "    S = np.abs(librosa.stft(y, n_fft=512)) \n",
    "    # https://librosa.org/doc/main/generated/librosa.stft.html (set 512 for speech processing)\n",
    "\n",
    "    # Extracting Features\n",
    "    tempo, beats = librosa.beat.beat_track(y=y, sr=sr)\n",
    "    chroma_stft = librosa.feature.chroma_stft(y=y, sr=sr, n_fft=512)\n",
    "    \n",
    "    chroma_cq = librosa.feature.chroma_cqt(y=y, sr=sr)\n",
    "    \n",
    "    chroma_cens = librosa.feature.chroma_cens(y=y, sr=sr)\n",
    "    melspectrogram = librosa.feature.melspectrogram(y=y, sr=sr, n_fft=512)\n",
    "    rmse = librosa.feature.rms(y=y)[0]\n",
    "    cent = librosa.feature.spectral_centroid(y=y, sr=sr, n_fft=512)\n",
    "    spec_bw = librosa.feature.spectral_bandwidth(y=y, sr=sr, n_fft=512)\n",
    "    contrast = librosa.feature.spectral_contrast(S=S, sr=sr, n_fft=512)\n",
    "    rolloff = librosa.feature.spectral_rolloff(y=y, sr=sr, n_fft=512)\n",
    "    poly_features = librosa.feature.poly_features(S=S, sr=sr, n_fft=512)\n",
    "    \n",
    "    tonnetz = librosa.feature.tonnetz(y=y, sr=sr)\n",
    "    \n",
    "    zcr = librosa.feature.zero_crossing_rate(y)\n",
    "    harmonic = librosa.effects.harmonic(y)\n",
    "    percussive = librosa.effects.percussive(y)\n",
    "\n",
    "    mfcc = librosa.feature.mfcc(y=y, sr=sr, n_fft=512)\n",
    "    mfcc_delta = librosa.feature.delta(mfcc)\n",
    "\n",
    "    onset_frames = librosa.onset.onset_detect(y=y, sr=sr)\n",
    "    frames_to_time = librosa.frames_to_time(onset_frames[:20], sr=sr)\n",
    "\n",
    "    # Concatenating Features into one csv and json format\n",
    "    feature_set['filename'] = filepath  # song name\n",
    "    feature_set['tempo'] = tempo  # tempo \n",
    "    feature_set['total_beats'] = sum(beats)  # beats\n",
    "    feature_set['average_beats'] = np.average(beats)\n",
    "    feature_set['chroma_stft_mean'] = np.mean(chroma_stft)  # chroma stft\n",
    "    feature_set['chroma_stft_var'] = np.var(chroma_stft)\n",
    "    \n",
    "    feature_set['chroma_cq_mean'] = np.mean(chroma_cq)  # chroma cq\n",
    "    feature_set['chroma_cq_var'] = np.var(chroma_cq)\n",
    "    \n",
    "    feature_set['chroma_cens_mean'] = np.mean(chroma_cens)  # chroma cens\n",
    "    feature_set['chroma_cens_var'] = np.var(chroma_cens)\n",
    "    feature_set['melspectrogram_mean'] = np.mean(melspectrogram)  # melspectrogram\n",
    "    feature_set['melspectrogram_var'] = np.var(melspectrogram)\n",
    "    feature_set['mfcc_mean'] = np.mean(mfcc)  # mfcc\n",
    "    feature_set['mfcc_var'] = np.var(mfcc)\n",
    "    feature_set['mfcc_delta_mean'] = np.mean(mfcc_delta)  # mfcc delta\n",
    "    feature_set['mfcc_delta_var'] = np.var(mfcc_delta)\n",
    "    feature_set['rmse_mean'] = np.mean(rmse)  # rmse\n",
    "    feature_set['rmse_var'] = np.var(rmse)\n",
    "    feature_set['cent_mean'] = np.mean(cent)  # cent\n",
    "    feature_set['cent_var'] = np.var(cent)\n",
    "    feature_set['spec_bw_mean'] = np.mean(spec_bw)  # spectral bandwidth\n",
    "    feature_set['spec_bw_var'] = np.var(spec_bw)\n",
    "    feature_set['contrast_mean'] = np.mean(contrast)  # contrast\n",
    "    feature_set['contrast_var'] = np.var(contrast)\n",
    "    feature_set['rolloff_mean'] = np.mean(rolloff)  # rolloff\n",
    "    feature_set['rolloff_var'] = np.mean(rolloff)\n",
    "    feature_set['poly_mean'] = np.mean(poly_features)  # poly features\n",
    "    feature_set['poly_var'] = np.mean(poly_features)\n",
    "    \n",
    "    feature_set['tonnetz_mean'] = np.mean(tonnetz)  # tonnetz\n",
    "    feature_set['tonnetz_var'] = np.var(tonnetz)\n",
    "    \n",
    "    feature_set['zcr_mean'] = np.mean(zcr)  # zero crossing rate\n",
    "    feature_set['zcr_var'] = np.var(zcr)\n",
    "    feature_set['harm_mean'] = np.mean(harmonic)  # harmonic\n",
    "    feature_set['harm_var'] = np.var(harmonic)\n",
    "    feature_set['perc_mean'] = np.mean(percussive)  # percussive\n",
    "    feature_set['perc_var'] = np.var(percussive)\n",
    "    feature_set['frame_mean'] = np.mean(frames_to_time)  # frames\n",
    "    feature_set['frame_var'] = np.var(frames_to_time)\n",
    "    \n",
    "    for ix, coeff in enumerate(mfcc):\n",
    "        feature_set['mfcc' + str(ix) + '_mean'] = coeff.mean()\n",
    "        feature_set['mfcc' + str(ix) + '_var'] = coeff.var()\n",
    "    \n",
    "    return feature_set\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "id": "86-2tFFt7PNS"
   },
   "outputs": [],
   "source": [
    "\n",
    "new_features_dict = extract_features('record.wav')\n",
    "df = pd.DataFrame([new_features_dict])\n",
    "df.to_csv('./new_record.csv', index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "colab": {
   "collapsed_sections": [],
   "provenance": []
  },
  "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": "3.11.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}