Kruti1234/mindcare-audio-emotion

MindCare Audio Emotion Recognition Model

This model detects emotions from speech audio.

Model Details

• Task: Audio Emotion Recognition
• Dataset: RAVDESS
• Architecture: CNN
• Accuracy: 91.32%
• Emotions: 8 classes

Emotions Detected

angry, calm, disgust, fearful, happy, neutral, sad, surprised

Usage

from huggingface_hub import hf_hub_download
import tensorflow as tf
import pickle

# Download model
model_path = hf_hub_download(repo_id="Kruti1234/mindcare-audio-emotion", filename="mindcare_audio_emotion_model.h5")
model = tf.keras.models.load_model(model_path)

# Download label encoder
encoder_path = hf_hub_download(repo_id="Kruti1234/mindcare-audio-emotion", filename="audio_label_encoder.pkl")
with open(encoder_path, 'rb') as f:
    label_encoder = pickle.load(f)

Model Info

• Input: MFCC features (40, 174, 1)
• Output: 7 emotion probabilities
• Framework: TensorFlow/Keras

Training

Trained on RAVDESS dataset with data augmentation.

License

Apache 2.0