Enriching Song Recommendation Through Facial Expression Using Deep Learning

• Published in: Ingénierie des systèmes d'Information Vol. 28; no. 1; pp. 225 - 229
• Main Author: Deore, Shalaka Prasad
• Format: Journal Article
• Language: English
• Published: Edmonton International Information and Engineering Technology Association (IIETA) 01.02.2023
• Subjects: Accuracy; Algorithms; Artificial neural networks; Computer vision; Datasets; Deep learning; Dictionaries; Emotion recognition; Emotional factors; Emotions; Listening; Literature reviews; Machine learning; Music; Musical performances; Musicians & conductors; Neural networks; Preferences; Recommender systems; Singers
• DOI: 10.18280/isi.280126
• ISSN: 1633-1311

The music recommendation systems are highly linked with the emotional response of the user as the majority of the music is based on the mood of the listener. A large number of researches have been performed for the detection of emotion through the use of a variety of different techniques. These approaches have been helpful in achieving the emotion of the subject using various devices and other hardware which can be highly expensive with very low rates of accuracy. Whereas the detection of expression of the subject can be useful in determining the mood or the emotion with a considerable degree of accuracy. Therefore, to achieve the effective identification of emotion of an individual for effective music recommendation has been proposed in this research paper. The presented approach utilizes image normalization and Convolutional Neural Networks (CNN) which are trained on a dataset consisting of a number of different emotional responses. This trained model is then used to determine the mood of the individual and recommend music based on the detected mood. The experimental evaluation of the approach is performed to determine the accuracy of the emotion recognition which has resulted in highly accurate results. We achieved 62.88% testing accuracy with MSE and RMSE values of 8.5 and 2.9 respectively. The obtained results are promising and show that the fuzzy classification technique optimizes the outcomes.