ashutoshroy02/hybrid-wave2vec-LSTM-emotion-stress-RAVDESS

Hybrid Wav2Vec2 + LSTM Emotion & Stress Model

This repository contains a trained hybrid Wav2Vec2 + LSTM model for emotion and stress recognition from raw speech audio.

Tasks

• Emotion classification: neutral, happy, sad, angry, fear
• Stress prediction: continuous score

Input

• .wav audio (16 kHz, mono)

Output

• Emotion label
• Stress score

Training Data

• RAVDESS

Inference and deployment code are included.