AudioStream.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "f257328f",
   "metadata": {},
   "source": [
    "This notebook is a simple example how to use DeGirum PySDK to do AI inference of an audio stream from local microphone. PyAudio is required to run this sample."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "577caf9a",
   "metadata": {},
   "outputs": [],
   "source": [
    "import sys\n",
    "import degirum as dg  # import DeGirum PySDK\n",
    "import numpy as np\n",
    "from IPython.display import clear_output"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "83ee6c6d",
   "metadata": {},
   "outputs": [],
   "source": [
    "# connect to model zoo\n",
    "zoo = dg.connect_model_zoo()\n",
    "\n",
    "# load model\n",
    "model = zoo.load_model(\"mobilenet_v1_yamnet_sound_cls--96x64_quant_n2x_orca_1\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "3ee28b09",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define model-specific audio streaming function\n",
    "# TL;DR: \n",
    "# We define context manager function, which opens PyAudio stream on enter, reads it and yields audio waveforms\n",
    "# of proper type, proper size, and with proper overlap. It properly closes PyAudio stream on exit.\n",
    "# You pass model parameters and check-for-abort function as arguments.\n",
    "\n",
    "import pyaudio\n",
    "from contextlib import contextmanager\n",
    "\n",
    "@contextmanager \n",
    "def AudioStream( model_info, check_abort ):\n",
    "    chunk_length = model_info.InputWaveformSize[0] // 2\n",
    "    audio = pyaudio.PyAudio()\n",
    "    stream = audio.open(format = pyaudio.paInt16, channels = 1,\n",
    "            rate = int(model_info.InputSamplingRate[0]), input = True, frames_per_buffer = chunk_length)\n",
    "    \n",
    "    data = np.zeros(2 * chunk_length, dtype = np.int16)\n",
    "    try:\n",
    "        def out_stream():\n",
    "            while not check_abort():\n",
    "                data[:chunk_length] = data[chunk_length:]\n",
    "                data[chunk_length:] = np.frombuffer(stream.read(chunk_length), dtype = np.int16)\n",
    "                yield data\n",
    "        yield out_stream\n",
    "    finally:\n",
    "        stream.stop_stream() # stop audio streaming\n",
    "        stream.close() # close audio stream\n",
    "        audio.terminate() # terminate audio library\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "db989e10",
   "metadata": {},
   "outputs": [],
   "source": [
    "abort = False # stream abort flag\n",
    "N = 5 # inference results history depth\n",
    "history = [] # list of N consecutive inference results\n",
    "\n",
    "# Acquire model input stream object\n",
    "with AudioStream(model.model_info, lambda: abort) as stream:\n",
    "    #\n",
    "    # AI prediction loop.\n",
    "    # emit keyboard typing sound to stop\n",
    "    #\n",
    "    for res in model.predict_batch(stream()):\n",
    "        # clear Jupyter output cell\n",
    "        clear_output(wait = True) \n",
    "        \n",
    "        # add top inference result to history\n",
    "        history.insert(0, f\"{res.results[0]['label']}: {res.results[0]['score']}\" )\n",
    "    \n",
    "        # keep only N last elements in history\n",
    "        if len(history) > N:\n",
    "            history.pop()\n",
    "    \n",
    "        # print history\n",
    "        for m in history:\n",
    "            print(m)\n",
    "        \n",
    "        # check for stop condition\n",
    "        if res.results[0]['label'] == \"Typing\":\n",
    "            abort = True\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "15d58c20",
   "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": "3.8.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}