This project aims to recognize emotions from audio signals using deep learning techniques. It processes .wav audio files, extracts meaningful features like MFCCs (Mel Frequency Cepstral Coefficients), and uses a neural network to classify the emotional tone in speech.

The notebook walks through data loading, preprocessing, model training, evaluation, and performance visualization.

├── isakovshohrukh-audioemotion.ipynb # Main Jupyter Notebook for training and evaluating the model
├── final_model.h5       # Trained TensorFlow model
├── final_data.csv       # Final dataset with extracted features
├── requirements.txt     # Project dependencies
├── README.md            # Project overview and instructions

• Audio preprocessing with librosa
• MFCC feature extraction
• Label encoding for emotion classes
• Neural network (MLP) for classification
• Accuracy tracking and confusion matrix visualization

Supported emotion labels include:

['neutral', 'calm', 'happy', 'sad', 'angry', 'fearful', 'disgust', 'surprised']

Install required libraries via pip:

pip install numpy pandas librosa matplotlib scikit-learn tensorflow

Or use a requirements.txt file:

numpy
pandas
librosa
matplotlib
scikit-learn
tensorflow

• Each .wav file contains a speech clip expressed with a specific emotion.
• The dataset should follow a naming convention or directory structure from which emotion labels can be parsed.