A Python-based real-time Speech Emotion Recognition project that listens to your voice through your microphone and predicts your current emotion (e.g., happy, sad, angry, etc.) using MFCC features and an MLP classifier.
The full project is hosted here: speech-emotion-detector
This project uses the RAVDESS (Ryerson Audio-Visual Database of Emotional Speech and Song) dataset, a validated multimodal collection of emotional speech and song performed by 24 professional actors across several emotions (calm, happy, sad, angry, fearful, surprise, disgust, neutral). Each expression is available in multiple modalities at different intensities and can be downloaded under a Creative Commons license from Zenodo :contentReference[oaicite:0]{index=0}.
- Trains on the RAVDESS dataset (speech-only, Emotion_1.zip)
- Extracts MFCC audio features (standardized to fixed-length)
- Achieves ~78% accuracy with an MLP classifier
- Real-time emotion prediction from your microphone input
- Built with Python, Librosa, scikit-learn, sounddevice, etc.
git clone https://github.com/your-username/speech-emotion-detector.git
cd speech-emotion-detector