问用Librosa从blob中提取特征

EN

{% extends 'base.html' %}

{% load static %}

{% block title %}Voice Detector{% endblock %}

{% block extracss %}
<link href="{% static 'css/voice_detector.css' %}" rel="stylesheet" />
<link href="{% static 'css/video-js.css' %}" rel="stylesheet" />
<link href="{% static 'css/all.min.css' %}" rel="stylesheet" />
<link href="{% static 'css/videojs.wavesurfer.min.css' %}" rel="stylesheet" />
<link href="{% static 'css/videojs.record.css' %}" rel="stylesheet" />
{% endblock %}

{% block content %}
<div class="banner">
  <div class="max-width">
    <div class="banner-content">
      <p class="motto">Test your voice right now!</p>
      <p class="description">
        Register your voice while reading the text below and our program will
        detect your gender in a few seconds!
      </p>
    </div>
  </div>
</div>
<div class="details">
  <section class="section">
    <div class="container">
      <div class="columns">
        <div class="column is-offset-4 is-4">
          <h1 class="title">Record audio</h1>
          <article class="message is-success" id="alert">
            <div class="message-header">
              <p>Recorded successfully!</p>
              <button class="delete" aria-label="delete"></button>
            </div>
            <div class="message-body">
              You have successfully recorded your message. You can now click on
              the Submit button to post it.
            </div>
          </article>
          <div class="field">
            <div
              class="control has-icons-left has-icons-right"
              style="margin-top: 1rem"
            >
              <audio
                id="recordAudio"
                class="video-js vjs-default-skin"
              ></audio>
            </div>
            <div class="control" style="margin-top: 1rem">
              <button class="home-btn" id="submit" disabled>Submit</button>
            </div>
          </div>
        </div>
      </div>
    </div>
  </section>
</div>
{% endblock %}

{% block extrajs %}
<script src="{% static 'js/video.min.js' %}"></script>
<script src="{% static 'js/RecordRTC.js' %}"></script>
<script src="{% static 'js/adapter-latest.js' %}"></script>
<script src="{% static 'js/wavesurfer.js' %}"></script>
<script src="{% static 'js/wavesurfer.microphone.min.js' %}"></script>
<script src="{% static 'js/videojs.wavesurfer.min.js' %}"></script>
<script src="{% static 'js/videojs.record.min.js' %}"></script>
<script src="{% static 'js/browser-workaround.js' %}"></script>
<script>
    // First lets hide the message
document.getElementById("alert").style.display = "none";
// Next, declare the options that will passed into the recording constructor
const options = {
  controls: true,
  bigPlayButton: false,
  width: 600,
  height: 300,
  fluid: true, // this ensures that it's responsive
  plugins: {
    wavesurfer: {
      backend: "WebAudio",
      waveColor: "#f7fff7", // change the wave color here. Background color was set in the css above
      progressColor: "#ffe66d",
      displayMilliseconds: true,
      debug: true,
      cursorWidth: 1,
      hideScrollbar: true,
      plugins: [
        // enable microphone plugin
        WaveSurfer.microphone.create({
          bufferSize: 4096,
          numberOfInputChannels: 1,
          numberOfOutputChannels: 1,
          constraints: {
            video: false,
            audio: true,
          },
        }),
      ],
    },
    record: {
      audio: true, // only audio is turned on
      video: false, // you can turn this on as well if you prefer video recording.
      maxLength: 180, // how long do you want the recording?
      displayMilliseconds: true,
      debug: true,
    },
  },
};

// apply audio workarounds for certain browsers
applyAudioWorkaround();

// create player and pass the the audio id we created then
var player = videojs("recordAudio", options, function () {
  // print version information at startup
  var msg =
    "Using video.js " +
    videojs.VERSION +
    " with videojs-record " +
    videojs.getPluginVersion("record") +
    ", videojs-wavesurfer " +
    videojs.getPluginVersion("wavesurfer") +
    ", wavesurfer.js " +
    WaveSurfer.VERSION +
    " and recordrtc " +
    RecordRTC.version;
  videojs.log(msg);
});

// error handling
player.on("deviceError", function () {
  console.log("device error:", player.deviceErrorCode);
});

player.on("error", function (element, error) {
  console.error(error);
});

// user clicked the record button and started recording
player.on("startRecord", function () {
  console.log("started recording!");

  $("#submit").prop("disabled", true);
});

// user completed recording and stream is available
player.on("finishRecord", function () {
  const audioFile = player.recordedData;

  console.log("finished recording: ", audioFile);

  $("#submit").prop("disabled", false);
  document.getElementById("alert").style.display = "block";
});

// Give event listener to the submit button
$("#submit").on("click", function (event) {
  event.preventDefault();
  let btn = $(this);
  //   change the button text and disable it
  btn.html("Submitting...").prop("disabled", true).addClass("disable-btn");
  //   create a new File with the recordedData and its name
  const recordedFile = new File([player.recordedData], `test.wav`);
  //   initializes an empty FormData
  let data = new FormData();
  //   appends the recorded file and language value
  data.append("file", recordedFile);

  //   post url endpoint
  $.ajax({
    url: "{% url 'detector' %}",
    method: "POST",
    data: data,
    dataType: "json",
    success: function (response) {
      if (response.success) {
        document.getElementById("alert").style.display = "block";
        window.location.href = `${response.url}`;
      } else {
        btn.html("Error").prop("disabled", false);
      }
    },
    error: function (error) {
      console.error(error);
    },
    cache: false,
    processData: false,
    contentType: false,
  });
});
</script>
{% endblock %}

def post(self, request):
        f = request.FILES['file']

        with open('file.wav', 'wb+') as destination:
            for chunk in f.chunks():
                destination.write(chunk)

        with open('file.wav', 'rb') as file:
            file_content = file.read()
            audio = wave.open('test.wav', 'wb')
            audio.setnchannels(1)
            audio.setnframes(1)
            audio.setsampwidth(1)
            audio.setframerate(8000)
            audio.writeframes(file_content)
            audio.close()

        (prediction, probability) = predict('test.wav')

        context["prediction"] = prediction
        context["probability"] = probability*100

        os.remove(file_name)

        return render(request, self.template_name, context=context)

def predict(file_name):
    # construct the model
    model = create_model()

    # load the saved/trained weights
    model.load_weights('model.h5')

    # extract features and reshape it
    features = extract_feature(file_name, mel=True).reshape(1, -1)

    # predict the gender!
    male_prob = model.predict(features)[0][0]
    female_prob = 1 - male_prob

    gender = "male" if male_prob > female_prob else "female"

    if gender == "male":
        return (gender, male_prob)

    return (gender, female_prob)

def extract_feature(file_name, **kwargs):
    mfcc = kwargs.get("mfcc")
    chroma = kwargs.get("chroma")
    mel = kwargs.get("mel")
    contrast = kwargs.get("contrast")
    tonnetz = kwargs.get("tonnetz")
    X, sample_rate = librosa.core.load(file_name)

    if chroma or contrast:
        stft = np.abs(librosa.stft(X))

    result = np.array([])

    if mfcc:
        mfccs = np.mean(librosa.feature.mfcc(y=X, sr=sample_rate, n_mfcc=40).T, axis=0)
        result = np.hstack((result, mfccs))

    if chroma:
        chroma = np.mean(librosa.feature.chroma_stft(S=stft, sr=sample_rate).T,axis=0)
        result = np.hstack((result, chroma))

    if mel:
        mel = np.mean(librosa.feature.melspectrogram(X, sr=sample_rate).T,axis=0)
        result = np.hstack((result, mel))

    if contrast:
        contrast = np.mean(librosa.feature.spectral_contrast(S=stft, sr=sample_rate).T,axis=0)
        result = np.hstack((result, contrast))

    if tonnetz:
        tonnetz = np.mean(librosa.feature.tonnetz(y=librosa.effects.harmonic(X), sr=sample_rate).T,axis=0)
        result = np.hstack((result, tonnetz))

    return result