/AudioEmotion

Recognize Audio Emotion.

Primary LanguagePythonMIT LicenseMIT

AudioEmotion

Recognize Audio Emotion.

Eight emotions: neutral, calm, happy, sad, angry, fearful, disgust and surprised.

Article

音频情绪识别

Usage

Create "AudioData", "Output/logs", "Output/mlmodels" folders. Download RAVDESS dataset and put it in "AudioData" folder.

Turicreate

  1. Replace some Turicreate source files by "Source/Turicreate_Fix/*.py"
  2. Train model using turicreate_classifier.py.
  3. run "iOSSample/AudioEmotion.xcodeproj" on iOS 12(or higher). You can also run predict_audio_emotion.py on your Mac/PC.

Keras

  1. run extract_audio_feature.py
  2. run train_audio.py
  3. run predict_audio_emotion.py