Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{172913,
        author = {Carol Maria Dsilva},
        title = {Music Recommendation Based on Facial Emotion Recognition Using Convolutional Neural Networks (CNN) and Cosine Similarity},
        journal = {International Journal of Innovative Research in Technology},
        year = {2025},
        volume = {11},
        number = {9},
        pages = {1378-1383},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=172913},
        abstract = {Music recommendation systems often struggle to capture users’ emotional states, leading to suboptimal song suggestions. Traditional approaches rely on historical data, explicit ratings, or demographic data, which fail to dynamically adapt to a user’s current mood. Counteracting this limitation, this study proposes a Facial Emotion Recognition-Based Music Recommendation System, integrating Convolutional Neural Networks (CNNs) for emotion detection and cosine similarity for personalized music recommendation.
The system uses the FER2013 dataset for training a deep CNN model capable of recognizing seven primary emotions. Extracted emotional states are then mapped to a curated music dataset, where cosine similarity is used to recommend songs with matching valence and energy levels. The model’s performance is evaluated using accuracy, precision, recall, and F1-score, achieving a competitive accuracy of 71.58% in emotion detection. The increase of relevance and user satisfaction in comparison to traditional recommendation models provides proof for the hypothesis. The proposed system not only enhances real-time adaptability but also eliminates dependency on user history, making recommendations more contextually appropriate.
Key contributions of this research include: (1) an end-to-end pipeline integrating facial emotion recognition with real-time music recommendation, (2) a CNN-based emotion detection model optimized for real-world applications, and (3) a cosine similarity-based approach for enhancing music recommendations. The findings demonstrate the potential of emotion-aware music recommendations in improving user experience and personalization. Future work will focus on expanding the emotion spectrum and refining song-matching techniques to further enhance recommendation accuracy.},
        keywords = {Facial Emotion Recognition, Convolutional Neural Networks, Music Recommendation, Cosine Similarity, Deep Learning.},
        month = {February},
        }