{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "fc6afb60-fe4a-402a-9aa6-37a24d0f50aa",
   "metadata": {},
   "source": [
    "# Winer denoise audio filter"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "1c3bdd16-6716-47f9-8b14-01315fc76a76",
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import torch\n",
    "\n",
    "import cutcutcodec"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2e160ab9-38d1-488d-9e17-edccfacb629e",
   "metadata": {},
   "source": [
    "## Read, Denoise and Write\n",
    "\n",
    "In a Winer filter, the noise profile has to be **ergodic**."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "845debd0-2fe4-452f-92cb-64d5b2247bdd",
   "metadata": {},
   "outputs": [],
   "source": [
    "FILE = cutcutcodec.utils.get_project_root().parent / \"media\" / \"audio\" / \"winer.wav\"\n",
    "RATE = 44100\n",
    "\n",
    "# read the signal\n",
    "with cutcutcodec.read(FILE) as noisy_container:\n",
    "\n",
    "    # extract the noise\n",
    "    noise_container = noisy_container.apply_audio_subclip(25.0, 30.0)\n",
    "\n",
    "    # denoised with winer\n",
    "    denoised_container = (noise_container | noisy_container).apply_audio_wiener(level=0.95, band=4.0)\n",
    "\n",
    "    # plot signals\n",
    "    noisy_stream = noisy_container.out_streams[0]\n",
    "    denoised_stream = denoised_container.out_streams[0]\n",
    "    noisy_signal = noisy_stream.snapshot(0, RATE, int(RATE*noisy_stream.duration))\n",
    "    denoised_signal = denoised_stream.snapshot(0, RATE, int(RATE*denoised_stream.duration))\n",
    "\n",
    "    # save signal\n",
    "    cutcutcodec.write(denoised_container.out_streams, \"/tmp/denoised_winer.flac\", streams_settings=[{\"encodec\": \"flac\", \"rate\": RATE}])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "93b050cd-c2bb-4120-8517-f9ed6bfe124b",
   "metadata": {},
   "source": [
    "## Plot spectrograms"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "d673092d-e7ac-4e37-8765-9d9daf3345d0",
   "metadata": {},
   "outputs": [],
   "source": [
    "# display noisy frame\n",
    "plt.title(\"Spectogram of the noisy signal\")\n",
    "plt.xlabel(\"time (s)\")\n",
    "plt.ylabel(\"frequency (Hz)\")\n",
    "plt.specgram(noisy_signal[0], Fs=RATE, NFFT=2048)\n",
    "plt.show()\n",
    "\n",
    "# display denoised frame\n",
    "plt.title(\"Spectogram of the denoised signal\")\n",
    "plt.xlabel(\"time (s)\")\n",
    "plt.ylabel(\"frequency (Hz)\")\n",
    "plt.specgram(denoised_signal[0], Fs=RATE, NFFT=2048)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c886f14e-2366-4040-96d9-ef0651565dde",
   "metadata": {},
   "source": [
    "## Plot temporal signals"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "557051ce-7039-4910-8c2a-32ac6ab3561d",
   "metadata": {},
   "outputs": [],
   "source": [
    "# signal in temporal domains\n",
    "plt.title(\"Winer filtering in temporal domain.\")\n",
    "plt.plot(torch.arange(noisy_signal.shape[1]) / RATE, noisy_signal[0], label=\"noisy signal\")\n",
    "plt.plot(torch.arange(denoised_signal.shape[1]) / RATE, denoised_signal[0], label=\"denoised signal\")\n",
    "plt.xlabel(\"time (s)\")\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  }
 ],
 "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": "3.13.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}