audio-emotion-recognition

Audio Emotion Recognition

This model recognizes emotions in speech audio files. It can detect the following emotions:

• Angry
• Disgust
• Fear
• Happy
• Neutral
• Sad
• Surprise

Model Description

This model uses a two-stage approach:

1. Audio feature extraction using YAMNet
2. Emotion classification using a custom neural network

Usage

Installation

pip install tensorflow librosa huggingface_hub tensorflow_hub

Quick Start

from huggingface_hub import snapshot_download
import os
import sys
sys.path.append(snapshot_download(repo_id="vkushwahaa/audio-emotion-recognition"))

from emotion_recognizer import load_from_hf

# Load model
recognizer = load_from_hf()

# Predict emotion from audio file
result = recognizer.predict("path/to/audio.wav")

# Print results
print(f"Predicted emotion: {result['predicted_emotion']} (confidence: {result['confidence']:.2f})")

Command Line Usage

python emotion_recognizer.py path/to/audio.wav

Training

This model was trained on multiple datasets:

• CREMA-D
• RAVDESS
• SAVEE
• TESS

The model uses YAMNet embeddings as features and employs data augmentation techniques to improve robustness.