Abstract:
In today’s fast-paced world, stress and emotional disturbances significantly impact mental well
being. Music has long been recognized as a powerful stress reliever, but its effectiveness depends
on aligning with the listener’s current mood. Existing affect-based music recommendation systems
are limited, primarily identifying only four emotions and lacking real-time personalization. To
address this gap, TUNE MOODS introduces a framework for emotion-driven music playlist
generation that detects a broader range of seven emotions Anger, Disgust, Fear, Happiness,
Sadness, Surprise, and Neutral providing more precise and personalized music recommendations.
Additionally, it suggests engaging activities such as games, videos, exercises, and inspirational
quotes to help users transition to a positive emotional state.
At the core of TUNE MOODS is a Convolutional Neural Network (CNN) trained for facial
expression recognition using a webcam. The model architecture consists of convolutional layers
for feature extraction, pooling layers for dimensionality reduction, and fully connected layers for
classification. The FER-2013 dataset was used for training and validation, enabling real-time
emotion recognition, which is then mapped to a curated music playlist. Furthermore, the system
employs a structured mood transition strategy, guiding users from negative emotions to positive
ones through music and interactive activities.
Performance evaluation showed that the CNN model achieved a Training Accuracy of 77.73%,
Validation Accuracy of 66.44%, and an F1 Score of 66%, demonstrating strong emotion
classification capabilities. A confusion matrix confirmed its ability to distinguish emotions with
high precision. User engagement metrics and feedback further validated that personalized music
recommendations and suggested activities effectively aid in emotional regulation. These results
establish TUNE MOODS as an innovative and holistic approach to integrating emotion
recognition and music therapy for improved mental well-being.
