Audio Emotion Recognition Using Machine Learning | Deep Learning Research Project

Worked on building and training an audio emotion recognition system capable of classifying human emotions from speech signals. Leveraged deep learning and signal processing techniques to extract relevant features and identify emotional patterns in voice data.

Key contributions:

Used Librosa to extract key audio features such as MFCCs, chroma, and zero-crossing rate.

Trained models like CNN, LSTM, and traditional classifiers to detect emotions including happy, sad, angry, neutral, and fear.

Applied data augmentation and noise handling techniques to improve robustness.

Evaluated models using accuracy, F1-score, and confusion matrix to identify best-performing architecture.

Gained hands-on experience in working with RAVDESS and TESS datasets for emotion classification tasks.

This project deepened my skills in speech-based affective computing, deep learning, and real-world audio analysis, paving the way for developing human-centric AI applications.